NUPOS: A part of speech tag set for written English from Chaucer to the present

By Martin Mueller, November 2009

1 Introduction and Summary

The following is a description of NUPOS, a part-of-speech (POS) tag set designed to accommodate the major morphosyntactic features of written English from Chaucer to the present day. The description is written for an audience not familiar with POS tagging. NUPOS is part of an enterprise to make the results of such tagging useful to humanities scholars who are not professional linguists and have not considered its utility for a wide variety of applications beyond linguistics proper.

While the NUPOS tag set can be used with any tagger that can be trained, so far it has been used only with Morphadorner, an NLP suite developed by Phil Burns and used extensively in the MONK project. Some 2,000 texts from the 1500s to the late 1800s have been tagged with it.

2 What is POS tagging?

A part-of-speech tag set is a classification system that allows you to as- sign some grammatical description to each word occurrence in a text. This assignment can be done by hand or automatically. Typically you “train” an automatic tagger by giving it the results of a hand-tagged corpus. The tagger then applies to unknown text corpora what it “learned” from the training set. The “knowledge” of the automatic tagger may consist of a set of rules or of a statistical analysis of the results. Either way, a good tagger will provide accurate descriptions for 97 out of 100 words.

Why do you want to apply POS tagging to a text in the first place? Readers might well ask this question when the sees the tagging output of the opening of Emma, which might look like this:

Emma_name Woodhouse_name, handsome_adj, clever_adj, and_conj rich_adj

This tells you nothing you did not know before. But humans are very subtle decoders who bring an extraordinary amount of largely tacit knowledge to the task of making sense of the characters on the page. The computer, however, lacks this knowledge. If you want to take full advantage of the query potential of a machine readable text you must make explicit in it at least some of the rudiments of readerly knowledge. If you do so, you can quickly and accurately perform many operations that will be difficult or practicable for human readers to do. You cannot only extract a list of adjec tives (or other parts of speech), you can also identify syntactic fragments, such as the sequence of three adjectives. A variety of stylistic or thematic opportunities for inquiry open up with a POS-tagged text, especially if the tagging is carried out consistently across large text archives. Analyses of this kind are based on the guiding assumption that there often is an illuminating path from low-level linguistic phenomena to larger-scale thematic or structural conclusions.

3 The concept of the LemPos

If you want to use computers for the analysis of texts that differ in time, genre, regional or social stratification you want to be in a position where the surface form of any word occurrence can be mapped to a more abstract representation that allows algorithms to identify features one surface form shares with others. For many purposes, a satisfactory mapping will consist of the combination of a part of speech tag with the lemma or the look-up form of the word in a dictionary. I call that combination a LemPos. Here are some examples:

Surface form or spelling Lemma + POS tag or LemPos
vniuersities university_ng
vniuersities university_n
university’s university_ng
universities university_n

Human readers tacitly process the ways in which these spellings stand for the same or different forms. The machine is not that bright, but once it has been presented with the ‘explicitated’ LemPos it can perform many operations that humans could never do with comparable speed or accuracy.

It is clear from this very simple example that the mapping of a spelling to a LemPos depends on three distinct operations:

  1. the recognition of orthographic variance
  2. the identification of morphosyntactic features
  3. the identification of the lemma

When the NUPOS tag set is used with MorphAdorner, the text for human readers or sequence of words on the printed is supplemented with a machine-readable representation that explicitly articulates some data while ignoring others

4 About tag sets

POS tags carry some combination of morphological and syntactic pieces of information, whence they are also called morphosyntactic tags. In highly inflected languages, such as Greek, Latin, or Old English, the inspection of a word out of context will reveal much about its grammatical properties. English has shed most of its inflectional features over the centuries, and the individual word will contain ambiguities that only context can resolve. Thus the –ed form of a verb may be the past tense or the past participle. For some common verbs (put, shut, cut), the distinction between past and present is morphologically unmarked. In many cases even the distinction between verb and noun (‘love’) is not morphologically marked.

In English, therefore, POS tagging is a business that works with very limited morphological information (mainly the suffixes –s, -ed, -ing, -er, -est, -ly) and uses the context of preceding or following words to make sense of things. A little reflection on these facts opens one’s eyes to characteristic errors of English taggers, such as the confusion of participial and past tense forms.

The most widely most used tag set for modern English is the Penn Tree- bank tag set. This set consists of about three dozen tags (though some of them can be combined). It offers a very crude classification system, but for many purposes it is good enough. When you are in the world of machines making decisions, crude distinctions consistently applied are more useful than error-ridden subtle distinctions. Like other modern tag sets, the Penn Treebank set lacks important feature for the accurate tagging of written English before the twentieth century. It recognizes the third person singular of a verb (VBZ), but it does not recognize the second person singular (‘thou art’). You can see the reason: the second person singular is no longer a living form. But it remains a living archaism, and it was a living form of poetic and religious usage well into the twentieth century.

Modern English taggers have a very odd way of dealing with the possessive case or genitive. In English orthography since the eighteenth century, the apostrophe has been used to distinguish between the –s suffix as a plural marker and as a possessive marker. Before the middle of the seventeenth century, this orthographical distinction is rarely or never found, and a sequence like “the kings command” is ambiguous.

The Penn Treebank set, like most other tag sets, treats the apostrophized ‘s’ as a separate word. When the automatic tagger applies its rules, a word like “king’s” is ‘tokenized’ as two words. The convenience of this procedure for modern English is obvious, especially since the apostrophized ‘s’ can also stand for ‘is’ or ‘has’ in contracted forms, where it has a linguistically sounder claim to be treated as a separate word. But if you want a tag set capable of processing written English across many centuries, it is clearly preferable to find a solution that treats the ‘s’ of the possessive case in the same way in which it treats other inflectional suffixes, such as the plural ‘s’ or the ‘ed’ and ‘ing’ of verb forms.

Like other English tag sets, the Penn Treebank set consists of a somewhat inconsistent mix of syntactic and morphological markers. The tags VVZ and NN2 respectively stand for the –s forms of a verb and a noun. In each case the symbol includes information about a syntactic category (verb, noun) and a morphological condition (3rd singular, plural). But the same morphological form can operate in different syntactic environment. This is particularly true of participial forms. When a form like ‘loving’ is used as a verb form, the code ‘VVG” provides information both about its syntactic function (VV) and its morphological form (G). But when the same word is used as an adjective or as a noun (the gerund), the codes JJ and NN ignore morphological information.

5 The NUPOS tag set

5.1 The history of the NUPOS tag set

The NUPOS tag set is a hybrid product that grew out of WordHoard, a project to create a search environment for deeply tagged corpora and includes all of Early Greek epic as well as the works of Chaucer, Spenser, and Shakespeare (https://wordhoard.northwestern.edu). The Greek texts were morphologically tagged with the help of the Morpheus tagger of the Perseus project. The Chaucer text was based on Larry Benson’s Glossarial Database to the Riverside Chaucer and uses the tag set designed by Benson for that project. The Shakespeare text was tagged with the CLAWS tag set developed at Lancaster University and used for the tagging of the British National Corpus.

My original plan was to use different tag sets for Chaucer and Shakespeare. But on closer inspection I discovered that you could with hardly any loss merge the Benson and CLAWS tags in a common set. It also turned out that Chaucer has only two verb forms that are not found in Shakespeare: the fairly rare second person plural imperative and the quite common –n form to mark the infinitive or first and third plural present of verbs.

In other words, you need only four tags to extend a modern tag set so that it can capture the major morphosyntactic phenomena in English from Chaucer on:

  1. The second person singular present
  2. The second person singular past
  3. The first and third plural present
  4. The second plural imperative

In merging the tag sets I took from Benson a “used-as” category that is important to his scheme and compensates for a weakness in the CLAWS and Penn Treebank sets. A word will typically belong to one word class and is used in all or most cases as an instance of that class. A noun is a noun, a verb is a verb, etc. But in a phrase like “no ifs or buts” the conjunctions ‘if’ and ‘but’ are used as nouns. In the catachrestic spirit of such a phrase you can use any word class as any other word class, and much word play de- pends on it.

There are more systemic uses of this phenomenon. In a phrase like ‘My loving lord’ the present participle of the verb ‘love’ is used as an adjective. In ‘the running of the deer’ a present participle is used as a noun. Benson’s tagging scheme explicitly recognizes these phenomena by creating code points like ‘present participle used as adjective’. This seems to me preferable to the practice of dropping the morphological information and using JJ or NN tags, as CLAWS and the Penn Treebank set do. The utility of keeping the information is particularly apparent if you are also lemmatizing a text and want to record adjectival uses of ‘loving’ or ‘loved’ as instances of the verb ‘love’.

The difficulties of classifying participial forms are worth some comment. English and its cognate languages distinguish sharply between nouns and verbs. They share number, but nouns lack voice and tense while verbs lack case and gender. But participles cross that divide. There are uses where a verbal, nominal, or attributive function clearly dominates, but there are many uses where it does not. The training data for participial forms in NUPOS follow the rule: “If in doubt it’s a verbal form.”

5.2 The structure of the NUPOS tag set

NUPOS owes some features to the morphological tagging scheme used in The Chicago Homer (www.library.northwestern.edu/homer). That scheme is taken over from Perseus’ Morpheus but it stores the information in a very atomic fashion in a relational database so that a given word can be retrieved as an instance of any of its grammatical properties, separately or in combination.

A Greek word can be adequately defined through the categories of tense, mood, voice, case, gender, person, number, degree. In conventional grammars, a description will typically consist of a string of properties, such as aor-ind-act-3rd-sing for the Greek word ‘eperse’. The VVZ tag of English tag sets does pretty much the same thing, but the ‘Z’ component implicitly specifies tense (present), person (3rd), and number (singular). If you keep the morphological information in a rigorously atomic and explicit fashion, you can search at different levels at granularity. For instance, any given in- stance of an aorist optative passive form in Greek will have person and number, but if you keep the information in what database experts call a ‘normalized’ fashion, you can ignore person and number (or any other atomic component) in your search.

The NUPOS tag set is implemented in a framework that supports the normalized representation of tag sets for different languages. A given form is defined by the values it holds in the categories of tense, mood, voice, case, gender, person, number, degree, wordclass and subclass, and part of speech. The categories of voice and gender are irrelevant to English, but you need both for Greek or Latin, and you need gender for French or German.

In assigning values to categories, I have made some practical decisions that may raise the linguists’ eyebrows. English has a residual subjunctive (If I were…), but no tagging scheme tries to recognize it, probably because it cannot be captured with sufficient accuracy by algorithms. My mood category quite properly includes the indicative and the infinitive. Somewhat less properly, it includes participles. In the ancient and modern European languages, participles may have voice or tense, but they lack mood and may therefore be put in a ‘mood’ column of a database without causing damage.

5.3 Negative forms and un-words

English has some contracted forms like ‘nas’ (was not), ‘niltow’ (ne wilt thou) or “don’t” whose orthographical status clearly testifies to their perception as single lexemes. If the subjunctive and optative moods are seen as modifications of the declarative indicative, why not accept a ‘negative’ form as a radical modification? The OED does something like it. If you look up ‘cannot’ you are told that it is “the ordinary modern way of writing can not.” But if you look at ‘can’ you are taken to its inflexions, where ‘cannot’ is described as the negative form of can. NUPOS adds a negative category that is used to discriminate between ‘will’ and “won’t”, ‘none’ and ‘one’, or ‘ever’ and ‘never’.

I have done something similar and perhaps more radical with ‘un-words’. Do ‘unforgiving’ and ‘unforgiven’ share a common lemma? If you decide to treat ‘un-’ words as negative forms, the question is easy to answer, and there are very clear rules for creating ‘un’ forms of English lemmata. Accordingly, I have treated the prefix ‘un-‘ as a negative modifier of a positive lemma, and its part of speech is given a -u flag. Thus ‘unnatural_j-u’ corresponds to ‘natural-j’.

There are always slippery cases. Since ‘do’ is put in the class of auxiliary verbs and the tagging does not distinguish between ordinary and auxiliary forms of the verb, the forms of ‘undo’ are not classified as forms of ‘do’, but its pos tags are given a -u flag anyhow, so that a search for -u forms will retrieve them.

If you reduce ‘un-words’ to their roots why not do the same thing for other prefixes, such as ‘under’ or ‘over’? There are two reasons for this. First, un- is by far the most common prefix. Secondly, un-words have a relatively weak status as stable lemmata in their own right. The modal case of an un-word is a participial adjective or adverb (unseen, undoubtedly), while the forms of verbs beginning with ‘over’ or ‘under’ are distributed much more evenly across infinitive, present, past, and participial forms.

5.4 Comparative and superlative forms

The comparative and superlative forms of adjectives are formed with the suffixes -er and -est for short adjectives and with the periphrastic forms ‘more’ and ‘most’ for long adjectives. I have classified ‘more’, ‘most’, ‘less’, ‘least’ as comparative and superlatives determiners with -c and -s flags so that a search for pos tags with those flags will let you measure the extent of comparative and superlative markers in a text.

5.5 Word Class and POS

The word class specifies the class to which a word belongs most of the time. The assignment is made on a lexical basis without reference to a particular context. There are major word classes, and some of them have subclasses. Taggers differ in their recognition of subclasses. NUPOS is more like CLAWS than the Penn Treebank tag set in recognizing subclasses. But you can ignore the subclasses if you wish.

The Penn Treebank tag set is very Spartan when it comes to verbs and does not distinguish between the open class of common verbs and the closed class of grammatical verbs. CLAWS recognizes modal verbs and has separate tags for each of the verbs ‘be’, ‘have’ and ‘do’. NUPOS follows CLAWS in this regard, largely because digitally assisted analysis increasingly makes use of syntactic fragments created by tag sequences, and in particular by tag trigrams. If you have any interest in such analysis you will want to distinguish between auxiliaries as markers of tense or voice: ‘had shot’ (vhd vvn) and ‘was shot’ (vbds vvn) are very different constructions.

Modal verbs present some problems of classification in a diachronic corpus. In Middle English, as in modern German, modal verbs are capable of ‘full’ uses: in both languages you can say things like “I can it not,” which you cannot do in modern English, just as you know cannot use ‘could’ as Chaucer used it in his description of the Wife of Bath:

Of remedies of love she knew per chaunce,
For she koude of that art the olde daunce.

Phrases of that kind are probably not uncommon in archaizing Early Modern English. NUPOS treats all forms of ‘may’, ‘will’, ‘shall’, ‘can’ and ‘ought’ as if they were modern modals, but it does recognize modal forms that are not possible in modern English, such as a modal participles or infinitives. Quasi-modals like ‘let’ and ‘used’ are treated as common verbs.

The modal verbs ‘can’, ‘will’, ‘may’, ‘shall’ each exist in two forms, which historically are present and past forms but in practice differ in mood rather than tense. It is worth marking the difference, because a discourse rich in ‘could, would, should’ is very different from a discourse rich in ‘can, will, shall’. It is easiest, and historically accurate, to mark it as a difference in tense.

5.6 POS or part of speech proper

The part-of-speech proper of any word occurrence is the syntactic role it plays in its context regardless of any particular morphological inflection. It is usually the same as the word class of a word, but in cases like ‘my loving lord’ it is not. The POS in this narrow sense is identical with the ‘used-as’ category in Benson’s tag set for Chaucer. It provides a very coarse classification of about two dozen categories, but for many purposes it may be good enough.

It is not easy to define the conditions that make you say: this noun (or verb) is not used as a noun (or verb) in this word occurrence. In compound nouns like ‘water closet’ the first noun acts as a kind of adjective; in a phrase like “the dead will rise” the adjective acts as a kind of noun. NUPOS assumes that such quasi-adjectival uses of nouns or quasi-nominal uses of adjectives are within the ordinary range of behaviour for nouns and adjectives. Therefore the POS for ‘water’ is noun and for ‘dead’ is adjective.

5.7 Ambiguous word classes

Some words cross word classes, and it is difficult for a computer program (or sometimes a human) to assign them confidently to a particular part of speech. Many of the mistakes that taggers make have to do with erroneous assignments of POS tags to such words. A particular occurrence of ‘since’ or ‘before’ may be an adverb, a preposition, or a conjunction. Many prepositions are used adverbially. The different uses of ‘as’ or ‘like’ are a nightmare to keep apart neatly.

NUPOS groups some words under the word class adverb-conjunction-preposition (ACP) and assigns its best guess to the POS tag. Thus an occurrence of ‘since’ may carry the tag C-ACP, which means “this is probably a conjunction but certainly an adverb, conjunction, or preposition.” Such a demarcation of the boundaries of error may be useful for some purposes. The terminology makes no special claim except that the classes of these words are likely to be confused with each other but not with other classes.

In addition to the ACP word class there are three other ambiguous word classes. Conjunctive, relative, and interrogative uses of the ‘wh- words’ are hard to tag automatically. I have bundled these words in a CRQ class, which includes such words as ‘who’, ‘which’, ‘when’, ‘why’ ‘what’.

Words like ‘yesterday’ or ‘today’ are largely adverbs, but have some nominal uses (yesterday’s paper). I have classified them as AN.

The last such class is a group of words that hover systematically between adjective and noun (JN). This class includes color words, names (Albanian, Jesuit, Florentine), and an odd assortment of words that include ‘evil’, ‘right’, ‘wrong’, ‘male’, ‘female’, ‘mercenary’ etc.

One could posit for each of these word a distinct lemma as noun and adjective, just as one distinguishes between the verb and the noun ‘love’. But I doubt whether ‘blue’ as noun or adjective is distinguished in the linguistic (un)conscious in the way in which the noun and verb ‘love’ are. It seems better to acknowledge that there is a class of words that systematically cross the boundaries of noun and adjective and whose properties can be described with some precision. The Oxford English Dictionary has it both ways with such words. Sometimes there are distinct entries, and sometimes you have an entry of the type “XX: adjective and noun.”

My criterion for classifying an adjective as a JN word has been its potential as a singular noun. You can say ‘my necessaries’ but not ‘my necessary’. But you can say ‘my secret’ or ‘a deep blue’. But these are very fluid distinctions. POS tagging is a very crude exercises and always reminds me of Wallace Stevens’ line from ‘Connoisseurs of Chaos’:

The squirming facts exceed the squamous mind

5.8 One word or many?

Automatic tagging of words relies on the normal case that a lexical unit consists of a single word separated by a space from the next word. The normal case is statistically more frequent than right-handedness. But there are a lot of ‘lefties’, and they pose a lot of challenges.

The lefties come in three forms. There are lexical units that span more than one word. There are hyphenated words, and there are contractions. Of these contractions pose the problem that is hardest to ignore because it forces you to make decisions about tokenization and POS assignment that do not in that form arise with multi word units or hyphenated forms. Although phrases like “according to” or “in vain” are most easily seen as instance of a two-word preposition or adverb, you can find ways of tagging each word separately. The component parts of a hyphenated word nearly always fit comfortably into an existing POS tag, most often an adjective or noun. But contracted forms typically cross the noun/verb divide and cannot be assigned to a single POS tag.

There are two different ways of approaching this problem, each with its own difficulties. In the first approach you say that contracted forms (much more common in speech than in writing) are “really” two words and that the written record should divide what lazy speaker slurred together. Alternately you can say that the orthographic practice of marking contractions, typically by means of the apostrophe, responds to a linguistic reality in the mind of the speakers or author and that the tagger ignores that reality when it keeps apart what the author intended to keep together.

For a variety of reasons, both practical and theoretical, NUPOS takes the second route. At the simplest level, you must “tokenize” words before you can apply POS tags to them. Tokenization has a number of consequences in a digital file. It counts the number of words and will play some role in as- signing to each word a unique address in a text. The closer the process of tokenization stays to the reader’s naïve perception the better off you are. Readers will say that in the sentence “Don’t do that” ‘that’ is the third word. You do not want to have to explain to them that it is the fourth word. Nor do you want to have a routine that counts it as the fourth word for some purpose and as the third word for others. Better to stick with the notion that “don’t do that” is a three-word sentence of which “don’t” is the first word.

Some contractions decompose easily into distinct parts, but others do not. Sometimes the apostrophe marks the division of words but sometimes it does not. In the case of “it’s” the apostrophe neatly divides the parts. In “’tis” or “don’t” the parts are easily identified, but the apostrophe is not the divider. In Early Modern English there are many contracted forms that are written as one word. ‘Nas’ for ‘ne was’ is one example. “Ain’t” is a modern example of a contracted form that is not easily decomposed, and it has as much right to be treated as a single token as ‘never’ or ‘none’.

Add these practical concerns to the assumption that the orthographic contraction reflects an underlying linguistic reality, and you come to the conclusion that contracted forms should be dealt with as single words as much as possible. That is the approach chosen in NUPOS.

The vast majority of contracted word occurrences—99% or more—are made up of a few very common patterns that are counted in the dozens rather than hundreds and amount to a closed class of combinations of pronouns and auxiliary/modal verbs or of auxiliary/modal verbs with the negative.

There is also an open class of verbs or nouns preceded by a contracted ‘to’ or ‘the’ (t’advance, th’earth) or a noun followed by the contracted form of ‘is’. You might call these proclitic and enclitic contractions.

If you treat a contracted form as a single word you still have to account separately for its components. As said above, combinations of an auxiliary or modal verb with a negative can be expressed in a single tag as the negative form of that verb. Combinations of a pronoun with an auxiliary or modal verb have to be expressed through a compound tag that joins the tag for the pronoun to the tag for the verb. Such compound tags raises the total number of tags (compound or single) by about a third.

Compound tags make life harder for the developer who designs the data object model and the interface for the user who formulates queries that depend on the tags for their answer. “She’ll” has to count for an instance of ‘will’ and ‘she.’ And the relevant form of ‘will’ in this case is “’ll” and not “she’ll.” Doing this in a consistent and user-friendly manner is not as easy as it sounds. But it is possible.

In Early Modern English, you find two-word spellings of forms that are now treated as single words. The most common cases are ‘to day’, ‘to morrow’ and reflexive pronouns like ‘myself’, ‘themselves’. MorphAdorner can and does tokenize these bigrams as single words so that a spelling like ‘them selues’ will appear in an XML representation of a text as

<w lemma="themselves" pos="pnx32">

5.9 The verb ‘be’

As in other languages, ‘be’ is the word with the largest and most diverse set of forms. Present tense forms include ‘art’, ‘is’, ‘are’, ‘be’, ‘be’st’ and ‘aren’. Past tense forms include ‘was’, ‘were’, ‘wast’, ‘wert’, and ‘weren’. There is only one form of the past participles, but it occurs in several orthographic variants.

In an earlier form of NUPOS, I mapped ‘is’ to ‘vbz’ and all other present forms to ‘vbb’. I mapped all the past forms to ‘vbd’. In this version, I use ‘vbr’ and ‘vbb’ to distinguish between ‘are’ and finite uses of ‘be’. I use ‘vbdr’ , ‘vbds’, ‘vbd2r’ and ‘vbd2s’ to distinguish between ‘were’, ‘was’, ‘wert’, and ‘wast’. These granular distinctions allow you to capture subtle distinctions between the forms. They also allow you to map variant spellings of the -r and -s form to standard spellings.

5.10 The ‘lempos’ and standardized spelling

With some exceptions and qualifications, the LemPos or combination of lemma and POS tag can be used to generate a standard spelling. You need an exception list of verbs and nouns that do not form their past and plural forms with -d or -s suffixes.

Adverbs pose a separate problem. The standard adverbial form of an adjective uses a -ly suffix. But there is a class of spatial adjectives that use an ‘- s’ suffix (‘downwards’). There is also a zero form of adverbs (‘pretty much’, ‘real soon’). The zero and - ly forms of some adjectives may have quite different meanings, as in the case of ‘just’, ‘very’, ‘pretty’, ‘straight’, or ‘hard’. Where there is strong semantic differentiation, it makes sense to split the adverb from its original lemma. Thus adverbial ‘hard’ and ‘hardly’, ‘just’ and ‘justly’, ‘very’ and ‘verily’ are treated as different lemmata.

You could solve this problem by having different tags for the zero, -s, and -ly forms of adverbs formed from adjectives.

Yet another problem is posed by variants that hover between morphological and orthographic variance – ‘loveth’ vs. ‘loves’ or ‘spake’ vs. ‘spoke’. Mapping ‘loveth’ to ‘loves’ or ‘spake’ to ‘spoke’ is less violent than mapping ‘wast’ to ‘wert’, but it does erase some real differences, as opposed to mapping ‘vniuersitie’ to ‘university’, where the differences are merely and systematically orthographic.

There are problems with homonyms. Depending on the meaning of the verb, the lempos ‘lie_vvd’ maps to the spellings ‘lay’ or ‘lied’. ‘Hanged’ and ‘hung’ are participial forms with quite distinct meanings, but they are both correctly described by the lempos ‘hang_vvd’.

You can go on with the enumeration of such problems. Some of them could in principle be resolved by more granular tag sets. Others resist algorithmic treatment. But it is also true that for the vast majority of cases, a LemPos can be mapped algorithmically to a single standard spelling.

5.11 How many tags and how many errors?

A good modern tagger will tag ~97% of words correctly. This is less impressive than it sounds because you can determine the part of speech of ~90% of all word occurrences from their lexical status. So from one perspective, the POS tagger makes a difference only for the last 10%, and it makes mistakes in a third of the cases.

Mistakes come in different shapes, and some matter more than others. For instance, the infinitive and present form of the verb are morphologically in- distinct. The infinitive is identified from a preceding ‘to’ or auxiliary verb. If other words intervene between the auxiliary and the verb mistakes are likely. Of 100 verb forms that are identified as VVB or VVI between 10 and 12 are likely to be classified wrongly. Perhaps wisely the Penn Treebank tag set does not even make the distinction. CLAWS and NUPOS try to make it because an infinitive always depends on another verb, and if you can exclude infinitive verbs from your count it is easier to count clauses. But for many users VVB/VVI errors are insignificant.

Another source of error is the confusion of the past participle (VVN) and the past tense (VVN). These too are morphologically indistinct except for a limited number of ‘strong’ verbs. In both NUPOS and CLAWS (at least when used with 16h century texts for which it was not designed) this error is more common than the confusion of VVB and VVI and may run as high as 15%-18%. If a form is correctly classified as a present or past participle its use may be incorrectly classified as a noun or an adjective.

Taggers using NUPOS will have trouble with identifying the possessive case of nouns where there is no apostrophe to mark it. Phrases like “the kings command” are genuinely difficult, and they involve a double error. The first mistake, classifying a possessive singular as a plural, is relatively benign. But if the tagger gets the first word wrong it may well make a mistake with the next word and classify a noun as a verb. That is a more consequential error: ng1-n1 is a very different syntactic construction from n2-vvb.

The coarser the classification, the lower the error rate. If you are satisfied with a broad classification of word occurrences as nouns, verbs, or adjectives, and do not worry about confusions of the VVB/VVI or VVD/VVN kind, the error rate probably drops by half.

5.12 Tagging at different levels of granularity

NUPOS is more explicit than other tagging schemes in letting users determine the granularity of the tagging. The NUPOS tag is really a “key” or unique ID that represents the classification of each morphological condition by discrete categories that users may ignore or activate. Depending on whether you classify by the strict POS tag, the combination of POS and wordclass, or the combination of all categories, you may end up with some twenty, sixty, or 250 tags.

6 Appendix

The following table shows the tag set for NUPOS. For each tag, the tag name is followed by an explanation, by an example, and by the approximate rate of occurrence per million words in 320 16th- and 17th-century English plays with a total word count of about six million words.

The NUPOS training data have included:

  1. The complete works of Chaucer and Shakespeare
  2. Spenser’s Faerie Queene
  3. North’s translation of Plutarch’s Lives
  4. Mary Wroth’s Urania
  5. Jane Austen’s Emma
  6. Dickens’ Bleak House and The Old Curiosity Shop
  7. Emily Bronte’s Wuthering Heights
  8. Thackeray’s Vanity Fair
  9. Mrs. Gaskell’s Mary Barton
  10. Frances Trollope’s Michael Armstrong
  11. George Eliot’s Adam Bede
  12. Scott’s Waverley
  13. Harriet Beecher Stowe’s Uncle Tom’s Cabin
  14. Melville’s Moby Dick

Examples are chosen for the most part from the training data.

NUPOS Tag set

NUPOS description example pos per million words
a-acp acp word as adverb I have not seen him since 6066.3
av adverb soon 35078.1
av-an noun-adverb as adverb go home 406.1
av-c comparative adverb sooner, rather 467.6
av-d determiner/adverb as adverb more slowly 1881.9
av-dc comparative determiner/adverb as adverb can less hide his love 1875.9
av-ds superlative determiner as adverb most often 931.7
av-dx negative determiner as adverb no more 854.2
av-j adjective as adverb quickly 8763.1
av-j-u adjective as adverb (un) unnaturally 90.2
av-jc comparative adjective as adverb he fared worse 731.7
av-jn adj/noun as adverb duly, right honourable 663.7
av-jn-u un-adj/noun as adverb (un-) unduly 0.3
av-jp proper adjective as adverb Christianly 0.5
av-jp-u proper adjective as adverb (un-) unchristianly 0.2
av-js superlative adjective as adverb in you it best lies 188.3
av-n noun as adverb had been cannibally given 0.2
av-s superlative adverb soonest 11.7
av-u adverb (un-) uneath 0.5
av-vvg present participle as adverb lovingly 76.9
av-vvg-u present participle as adverb (un-) unknowingly 1.4
av-vvn past participle as adverb Stands Macbeth thus amazedly 17.5
av-vvn-u past participle as adverb (un-) undoubtedly 6.6
av-x negative adverb never 1607.6
avc-jn comparative adj/noun as adverb deeper 8.0
avs-jn superlative adj/noun as adverb hee being the worthylest constant  
c-acp acp word as conjunction since I last saw him 8886.8
c-crq wh-word as conjunction when she saw 5271.7
cc coordinating conjunction and, or 32276.6
cc-acp acp word as coordinating conjunction but 6267.8
ccx negative conjunction nor 1234.6
crd numeral 2, two, ii 4378.3
cs subordinating conjunction if 8093.1
cst ‘that’ as conjunction I saw that it was hopeless 9263.7
d determiner that man, much money 28653.1
dc comparative determiner less money 946.4
dg determiner in possessive use the latter’s 4.6
dgx negative determiner in possessive use neither’s 0.3
ds superlative determiner most money 381.5
dt article a man, the man 49407.5
dx negative determiner as adverb no money 3185.9
fw-es Spanish word cuerpo 21.0
fw-fr French word monsieur 642.4
fw-ge German word Herr 104.4
fw-gr Greek word kurios 8.6
fw-it Italian word cambio 42.9
fw-la Latin word dominus 1662.9
fw-mi word in unspecified other language n/a 169.0
j adjective beautiful 43855.4
j-av adverb as adjective the then king 0
j-jn adjective-noun the sky is blue 5647.8
j-jn-u adjective-noun (un-) undue 24.6
j-u adjective (un-) unnatural 650.2
j-vvg present participle as adjective loving lord 1700.5
j-vvg-u present participle as adjective (un-) unrelenting spirit 34.1
j-vvn past participle as adjective changed circumstances 2260.8
j-vvn-u past participle as adjective (un-) unblemished night 489.2
jc comparative adjective handsomer 1457.1
jc-jn comparative adj/noun yet she much whiter 61.9
jc-u comparative adjective (un-) unhappier 0.3
jc -vvg present participles as comparative adjective for what pleasinger then varietie, or sweeter then flatterie? 0.2
jc-vvn past participle as comparative adjective shall find curster than she 0.7
jp proper adjective Athenian philosopher 916.9
jp-u proper adjective (un-) unchristian 1.2
js superlative adjective finest clothes 1472.5
js-jn superlative adj/noun reddest hue 163.4
js-jn-u superlative adj/noun (un-) unwelcomest man 0.3
js-n noun as superlative adjective felonest (Spenser)  
js-u superlative adjective (un-) unworthiest hand 4.7
js-vvg present participle as superlative adjective the lyingest knave in Christendom 6.4
js-vvn past participle as superlative adjective deformed’st creature 4.7
js-vvn-u past participle as superlative adjective (un-) the unprovidest sir of all our courtesies 0.2
n-jn adj/noun as noun a deep blue 1239.3
n-jn-u adj/noun as noun(un) through myn unkonninge (Chaucer) 0
n-vdg present participle as noun, ‘do’ my doing 2
n-vhg present participle as noun, ‘have’   0
n-vvg present participle as noun the running of the deer 862.9
n-vvg-u present participle as noun (un-) the clear unfolding of my doubts 9.7
n-vvn past participle as noun the departed 16.8
n1 singular, noun child 140905.8
n1-an noun-adverb as singular noun my home 169.5
n1-j adjective as singular noun an important good 0.2
n1-u singular, noun (un-) unthrift 64.9
n2 plural noun children 35795.9
n2-acp acp word as plural noun and many such-like “As’es” of great charge 0.2
n2-an noun-adverb as plural noun all our yesterdays 6.9
n2-av adverb as plural noun and are etcecteras no things 0.3
n2-cc coordinating conjunction used as noun and’s 0.3
n2-crq wh-word used as noun why’s 0.3
n2-dx determiner/adverb negative as plural noun yeas and honest kerysey noes 0.5
n2-j adjective as plural noun give me particulars 185.1
n2-jn adj/noun as plural noun the subjects of his substitute 669.2
n2-sy character used as plural noun her C’s 1.9
n2-u plural noun (un-) serious untruths 7.1
n2-uh interjection used as noun in russet yeas 0.8
n2-vdg present participle as plural noun, ‘do’ doings 9.8
n2-vhg present participle as plural noun, ‘have’ my present havings 0.3
n2-vvg present participle as plural noun the desperate languishings 164.1
n2-vvg-u present participle as plural noun (un-) undoings 0.2
n2-vvn past participle as plural noun there was no necessity of a Letter of Slains for Mutilation 0
ng1 singular possessive, noun child’s 3308.5
ng1-an noun-adverb in singular possessive use Tomorrow’s vengeance 1.7
ng1-j adjective as possessive noun the Eternal’s wrath 0.7
ng1-jn adj/noun as possessive noun our sovereign’s fall 45.1
ng1-vvn past participle as possessive noun knock at the closed door of the late lamented’s house 0.2
ng2 plural possessive, noun children’s 349.0
ng2-j adjective as plural possessive noun the poors’ cries 1.2
ng2-jc comparative adjective as possessive plural noun hindering the greaters’ growth 0.2
ng2-jn adj/noun as plural possessive noun mortals’ chiefest enemy 32.9
njp proper adjective as noun a Roman 57.6
njp proper adjective as plural noun The Romans 196.4
njpg proper adjective as possessive noun The Roman’s courage 7.6
njpg proper adjective as plural possessive noun The Romans’ courage 17.6
np1 singular, proper noun Paul 16703.6
np1-n singular noun as proper noun at the Porpentine 43.1
np2 plural, proper noun The Nevils are thy subjects 232.7
np2-n plural noun as proper noun such Brooks are welcome to me 0.3
npg singular possessive, proper noun Paul’s letter 1383.2
npg1-n singular possessive noun as proper noun and through Wall’s chink 3.2
npg plural possessive, proper noun will take the Nevils’ part 5.1
ord ordinal number fourth 1862.5
p-acp acp word as preposition to my brother 64612.9
pc-acp acp word as particle to do 14699.0
pi singular, indefinite pronoun one, something 1261.4
pi2 plural, indefinite pronoun from wicked ones 68.8
pi2x plural, indefinite pronoun To hear my nothings monstered 5.3
pig singular possessive, indefinite pronoun the pairings of one’s nail 12.2
pigx possessive case, indefinite pronoun nobody’s 0
pix indefinite pronoun none, nothing 1394.7
pn22 2nd person, personal pronoun you 18844.4
pn31 3rd singular, personal pronoun it 8254.1
png11 1st singular possessive, personal pronoun a book of mine 476.1
png12 1st plural possessive, personal pronoun this land of ours 78.8
png21 2nd singular possessive, personal pronoun this is thine  
png22 2nd person, possessive, personal pronoun this is yours 267.3
png31 3rd singular possessive, personal pronoun a cousin of his 304.4
png32 3rd plural possessive, personal pronoun this is theirs 30.3
pno11 1st singular objective, personal pronoun me 9589.0
pno12 1st plural objective, personal pronoun us 1904.1
pno21 2nd singular objective, personal pronoun thee 3070.5
pno31 3rd singular objective, personal pronoun him, her 7820.2
pno32 3rd plural objective, personal pronoun them 2560.3
pns11 1st singular subjective, personal pronoun I 26062.5
pns12 1st plural subjective, personal pronoun we 4069.0
pns21 2nd singular subjective, personal pronoun thou 4814.7
pns31 3rd singular subjective, personal pronoun he, she 9647.8
pns32 3rd plural objective, personal pronoun they 3104.9
po11 1st singular, possessive pronoun my 15833.9
po12 1st plural, possessive pronoun our 3379.5
po21 2nd singular, possessive pronoun thy 4370.3
po22 2nd person possessive pronoun your 9585.3
po31 3rd singular, possessive pronoun its, her, his 10050.7
po32 3rd plural, possessive pronoun their 2675.1
pp-f preposition ‘of’ of 18369.2
px11 1st singular reflexive pronoun myself 762.2
px12 1st plural reflexive pronoun ourselves 116.8
px21 2nd singular reflexive pronoun thyself, yourself 620.3
px22 2nd plural reflexive pronoun yourselves 89.5
px31 3rd singular reflexive pronoun herself, himself, itself 736.3
px32 3rd plural reflexive pronoun themselves 179.3
pxg21 2nd singular possessive, reflexive pronoun yourself’s remembrance 0.2
q-crq interrogative use, wh-word, subject Who? What? How? 5915.6
qg-crq interrogative use, wh-word, possessive Whose? 12.7
qo-crq interrogative use, wh-word, object Whom? 38.1
r-crq relative use, wh-word, subject the girl who ran 5601.9
rg-crq relative use, wh-word, possessive to such, whose faces are all zeal 782.0
ro-crq relative use, wh-word, object a wretched maid, whom ye have pursued 640.3
sy alphabetical or other symbol A, @ 233.6
uh interjection oh! 6484.7
uh-av adverb as interjection Well! 475.8
uh-crq wh-word as interjection Why, there were but four 827.5
uh-dx negative interjection No! 889.7
uh-j adjective as interjection Grumio, mum! 13.4
uh-jn adjective/noun as interjection And welcome, Somerset 82.5
uh-n noun as interjection Soldiers, adieu! 315.1
uh-np proper noun as interjection Jesu 0.2
uh-v verb as interjection My gracious silence, hail 155.4
uh-x negative interjection No! 843.6
vb2-imp 2nd plural present imperative, ‘be’ Beth pacient  
vb2r 2nd singular present of ‘be’ thou art 711.7
vb2rx 2nd singular present, ‘be’ thow nart yit blisful  
vb2s 2nd singular present of ‘be’ thou beest 23.6
vbb present tense, ‘be’ they be 2559.0
vbbx present tense negative, ‘be’ aren’t, ain’t, beant 0.5
vbd2r 2nd singular past of ‘be’ wert 93.6
vbd2s 2nd singular past of ‘be’ wast 32.7
vbd2x 2nd singular past, ‘be’ weren’t  
vbdp plural past tense, ‘be’ whose yuorie shoulders weren couered all  
vbdr past tense, ‘be’ were 1903.6
vbdrx past tense negative, ‘be’ weren’t, nere (Chaucer)  
vbds past tense, ‘be’ was 2588.5
vbdsx past tense negative, ‘be’ wasn’t, nas (Chaucer)  
vbg present participle, ‘be’ being 650.0
vbi infinitive, ‘be’ be 6414.1
vbm 1st singular, ‘be’ am 2705.1
vbmx 1st singular negative, ‘be’ I nam nat lief to gabbe 0.2
vbn past participle, ‘be’ been 999.7
vbp plural present, ‘be’ Thise arn the wordes 0.2
vbr present tense , ‘be’, ‘are’ they are 4674.2
vbrx present tense negative, ‘be’, are they aren’t 0.2
vbz 3rd singular present, ‘be’ is 8820.2
vbzx 3rd singular present negative, ‘be’ isn’t 0
vd2 2nd singular present of ‘do’ dost 431.5
vd2-imp 2nd plural present imperative, ‘do’ Dooth digne fruyt of Penitence 0
vd2x 2nd singular present negative, ‘do’ thee dostna know the pints of a woman 0.2
vdb present tense, ‘do’ do 3093.9
vdbx present tense negative, ‘do’ don’t 2.7
vdd past tense, ‘do’ did 1416.8
vdd2 2nd singular past of ‘do’ didst 155.3
vdd2x 2nd singular past negative, verb Why, thee thought’st Hetty war a ghost, didstna? 0
vddp plural past tense, ‘do’ on Job , whom that we diden wo 0
vddx past tense negative, ‘do’ didn’t 0
vdg present participle, ‘do’ doing 52.2
vdi infinitive, ‘do’ to do 1003.2
vdn past participle, ‘do’ done 766.3
vdp plural present, ‘do’ As freendes doon whan they been met 0
vdz 3rd singular present, ‘do’ does 1185.1
vdzx 3rd singular present negative, ‘do’ doesn’t 0
vh2 2nd singular present of ‘have’ thou hast 559.8
vh2-imp 2nd plural present imperative, ‘have’ O haveth of my deth pitee! 0
vh2x 2nd singular present negative, ‘have’ hastna 0
vhb present tense, ‘have’ have 5394.4
vhbx present tense negative, ‘have’ haven’t 4.2
vhd past tense, ‘have’ had 1821.0
vhd2 2nd singular past of ‘have’ thou hadst 92.4
vhdp plural past tense, ‘have’ Of folkes that hadden grete fames 0
vhdx past tense negative, ‘have’ hadn’t 0.2
vhg present participle, ‘have’ having 157.6
vhi infinitive, ‘have’ to have 2239.8
vhn past participle, ‘have’ had 155.1
vhp plural present, ‘have’ They han of us no jurisdiccioun 0
vhz 3rd singular present, ‘have’ has, hath 2753.6
vhzx 3rd singular present negative, ‘have’ Ther loveth noon, that she nath why to pleyne. 0
vm2 2nd singular present of modal verb wilt thou 921.7
vm2x 2nd singular present negative, modal verb O deth, allas, why nyltow do me deye 0
vmb present tense, modal verb can, may, shall, will 17429.8
vmb1 1st singular present, modal verb Chill not let go, zir, without vurther ‘cagion 0.7
vmbx present tense negative, modal verb cannot; won’t; I nyl nat lye 1039.8
vmd past tense, modal verb could, might, should, would 6475.3
vmd2 2nd singular past of modal verb couldst, shouldst, wouldst; how gret scorn woldestow han 264.2
vmd2x 2nd singular present, modal verb Why noldest thow han writen of Alceste 0
vmdp plural past tense, modal verb tho thinges ne scholden nat han ben doon. 0
vmdx past negative, modal verb couldn’t; She nolde do that vileynye or synne 1.2
vmi infinitive, modal verb Criseyde shal nought konne knowen me. 0
vmn past participle, modal verb I had oones or twyes ycould 0
vmp plural present tense, modal verb and how ye schullen usen hem 0
vv2 2nd singular present of verb thou knowest 975.6
vv2-imp 2nd present imperative, verb For, sire and dame, trusteth me right weel, 0
vv2-u 2nd singular present of verb (un-) thou unbendest 0.3
vv2x 2nd singular present negative, verb “Yee!” seyde he, “thow nost what thow menest; 0
vvb present tense, verb they live 38328.6
vvb-u present tense, verb (un-) they unfold 56.6
vvbx present tense negative, verb What shall I don? For certes, I not how 0.2
vvd past tense, verb knew 10730.8
vvd-u past tense, verb (un-) he unlocked the horse 7.3
vvd2 2nd singular past of verb knewest 159.5
vvd2-u 2nd singular past of verb (un-) thy treacherous blade unrippedest the bowels 0.2
vvd2x 2nd singular past negative, verb thou seidest that thou nystist nat  
vvdp past plural, verb They neuer strouen to be chiefe  
vvdx past tense negative, verb she caredna to gang into the stable  
vvg present participle, verb knowing 4715.1
vvg-u present participle, verb (un-) without unveiling herself 7.6
vvi infinitive, verb to know 44589.5
vvi-u infinitive, verb (un-) I must unclasp me 96.6
vvn past participle, verb known 20285.1
vvn-u past participle, verb (un-) would you be thus unclothed 147.5
vvp plural present, verb Those faytours little regarden their charge 1.0
vvp-u plural present, verb (un-) They unsowen the semes of freendshipe (Chaucer)  
vvz 3rd singular preseent, verb knows 10287.8
vvz-u 3rd singular preseent, verb he that unbuckles this 7.8
vvzx 3rd singular present negative, verb She caresna for Seth. 0
wd word wrongly split or joined in text   546.4
xx negative not   10210.2
zf English word wrongly used by foreign speaker   102.2
zz unknown or unparsable token   2312.4