Tag Archives: collocation

Under the surface: SHARP, LDNA and sundry sources

This blog post excerpts material Iona wrote reflecting back on her contribution to the SHARP conference in Paris in July 2016, building on the work of her PhD thesis and incorporating material and processes that have formed part of the Linguistic DNA project. The full post can be found on Iona’s personal blog.


In preparation for the paper, I dedicated time to manually extract, compile and refine measurements for some of the early outputs from the LDNA processor. To fit in with the pledges of my abstract, I targeted the associations of valour and valiant in subsets of EEBO-TCP.

During my PhD, I used EEBO-TCP to provide context for my work with early modern bibles. Valour entered the equation as I examined trends in the translation of a Hebrew collocation gibbor chayil. In the King James Version (publ. 1611) most gibbor chayil men are “mighty . . . of valour”. The repetition of this phrase across the translation means that English bible readers could form associations between the group of characters referred to, in a similar manner to those who encounter the Hebrew narrative directly. For this to happen in translation shows that the translators recognised and (sometimes) prioritised the transmission of this connection; in this respect “mighty of valour” is a partial example of a larger trend in favour of a more technical approach to translation, a move likely influenced by the increasing use of precise cross-referencing in bible reading (facilitated by the introduction of verse numbers throughout the Bible, an innovation of the 1550s). Yet the phrase is intrinsically interesting because before that “valour” was not part of the English biblical lexicon.

Collating instances of gibbor chayil demonstrates that the lexically related “valiant” was used in earlier translations, but in a piecemeal manner (illustrated by the changing distribution of black square bullets in the diagram below).

Poster

This diagram, extracted from my SHARP presentation, is one of a series colour-coded to highlight consistency within individual versions with a focus on the characterisation of Boaz. The black square bullets are added to highlight where a form of ‘valiant’ (or for KJ ‘valour’) was used.

By exploring the words valiant and valour with the LDNA tools, I was able to corroborate the impression I had formed during my earlier quantitative and qualitative analysis which was conducted via a standard EEBO-TCP interface.

The PhD bit

Searching hits in the population for the first century of English print (to 1570) and comparing that with the next half century (a collection of documents three times the size) I had observed that the frequency of both valiant and valour increased markedly above expectation.

Poster

Comparison of word frequency (hits) and distribution (records, hits per record) in EEBO-TCP for 1473-1570 (P1) and 1571-1620 (P2) expressed in ratios.

Scrutinising the data by decade exposed some significant textual influences. To quote from my thesis:

87 per cent of occurrences of “valiant” in the corpus for 1520-1529 (316 of a total 363) appear in a two-volume translation of the French chronicles of Froissart, while two other translated works account for a further 9 per cent; just 4 per cent of hits occur in ‘indigenous’ texts.

For “valour”,

a jump in the decade 1570-1579 is significantly related to the publication in 1579 of a translation from Italian: 403 of the decade’s 501 hits appear in a one-volume translation of The historie of Guicciardin conteining the vvarres of Italie and other partes (London, 1559). Once such scrutiny is imposed, it becomes evident that translation had a significant role in the increased currency of these two Latinate terms. It is also evident that the words normally appear in certain genres: conduct books concerned with warfare and chivalric behaviour; and chronicles of past history. This contributes to the recognisable sense of valour as “The quality of mind which enables a person to face danger with boldness or firmness; courage or bravery, esp. as shown in warfare or conflict; valiancy, prowess.”[ OED s.v. “valour|valor, n.”, §1c.] This sense, cultivated through translation in the course of the sixteenth-century, fits the context in which King James’ translators employ the word.

The LDNA bit

The subsets of EEBO-TCP sent through the LDNA processor earlier in the year were intentionally compatible with the periodisation of my thesis, providing windows onto English discourse that could be cross-referenced with the publication of particular bibles. The subsets thus incorporate all transcribed material from EEBO (TCP update 2015) known to have been printed during the following spans:

  • 1520-1539 (cf. Coverdale Bible 1535, Matthew Bible 1537, Great Bible 1539)
  • 1550-1559 (Geneva Bible 1560, Bishops Bible 1568); and
  • 1610-1611 (Douai Old Testament 1609-10, King James Version 1611).

Taking the first and last of these, measuring PMI in windows of discourse around the word “valour”, we find marked change in the prominent associations. Our approach yields plentiful data, and we are still thinking through the challenges of visualisation. In the slide shown, I have coloured associated terms according to the innermost window in which the cooccurring lemma rises to prominence. Thus red terms occur frequently in the narrowest window around valour (+/-1 words), orange terms in the expanded window (+/-10 words) that might approximate the surrounding sentence, green for +/-50 words (which now form the default window size in our public interface) and blue for the wide discursive window of +/-100 words. (Many lemmas appear in more than one window, and the list shown for the later period does not reach to some relevant low frequency items such as “prowess”.)

Poster

What should be visible is a distinction between the use of “valour” as a synonym of value or worth (prominent in the 1520-1539 subset), and the association with conduct in conflict (dominant in the 1610-1611 dataset). Both senses were part of the Latin root “valeo” and, had King James’ translators ventured it, both could have been played upon to make even more “mighty men of valour” in 1611. (One of the exceptions comes at 2 Kings 15:20, where Menachem taxes all gibbor chayil men, “mighty men of wealth” in the KJV.)

Inevitably, the set of observations I could draw from this investigation are not part of the bottom-up process that LDNA strives to achieve. But the exercise has helped me to think through some different ways we will want to be able to interrogate our data and to study the effects of some different baselines for our expectation calculations. And it demonstrates, I think, the valour of conducting semantic enquiries through discursive windows.

_____

Notes

Thesis quotations are from: I. C. Hine, “Englishing the Bible in early modern Europe: The case of Ruth”, PhD thesis (University of Sheffield, 2014), p. 163. These numbers reflect searches conducted through the Chadwyck EEBO interface using its variant spelling option.

The datasets employed in my thesis are not quite identical to those used by the project: LDNA uses a slightly expanded version of the EEBO-TCP collection (last updated early 2015) with its spelling regularised and tokens lemmatised locally using MorphAdorner.

Proximity Data

Background

The Linguistic DNA project will be interrogating cleaned-up EEBO and ECCO data in various ways, to get at its lexical semantic and conceptual content. But how do we get semantic and conceptual information from textual data? Sticking with  the original project proposal, we begin with an analysis of ‘proximity data’. What is proximity data, what does it tell us, and how can we measure it?

What is proximity?

Proximity relates to co-occurrence between terms in language. So, what is a term and what does it mean to co-occur?

A term may be:

  • a single word, a pair of words (or bigram), or a string of three or more words in order (an n-gram);
  • a grammatical construction whose ‘slots’ can be filled with appropriate words (e.g. ‘NOUN of NOUN’, ‘ADJECTIVE as NOUN’, or even ‘VERB MODIFIER DIRECT OBJECT’);
  • a phrase with lexical wild cards such as ‘very ___ ideas’.

Co-occurrence can then be defined as the presence of two or more terms within a given set of data, or in a given relationship. For example, we might be interested in the co-occurrence of two single words like Lord and law: In which texts do those terms co-occur? How close is one to the other? Or, we might be interested in the co-occurrence of a single word with a grammatical pattern: In which texts is see followed by a subordinate clause?

How do we investigate proximity?

We can ask a few different things about the distance between terms that co-occur. For example, we can inquire: ‘What terms occur within a given distance of term a (e.g. Lord)?’ Or, we can ask: ‘How far is term a (e.g. Lord) from term b (e.g. law)?’ Put differently, we can measure co-occurrence by selecting a starting point term (a node) and a distance from that starting point, and seeing what terms occur within that distance. Alternatively, we can select multiple nodes as starting points and measure the distance between them in use. We can also combine these two methods: we can first ask what words occur within a given distance of term a, and then take pairs of words from the resulting list and ask just how closely they occur to each other.

Finally, we can ask: ‘What occurs in a given relationship to term a?’ These questions can be syntactic: ‘What are the Direct Objects and Subjects of term a (e.g. see)?’ or related to Parts Of Speech (POS) ‘What noun occurs most frequently after term a (e.g. see)?’ We can also hypothetically ask about semantic relationships: ‘What is the Agent or Patient, Instrument or Theme related to term a?’ A syntactic approach is employed by the commercially-developed Sketch Engine software, and also generally, in various ways, in the Behavioural Profiling technique used by Stefan Gries (2012), in the collostructional approach used by Anatol Stefanowitsch and Gries (2008) and by Martin Hilpert (2012). This approach requires either satisfactory automated syntactic parsing or manual syntactic parsing—both of which seem to be impossible with EEBO because of the scale and variation documented previously. A POS approach is more viable with EEBO, but still difficult.

An alternative to syntactic and POS approaches is pair-pattern matrices: rather than investigating co-occurrence within grammatical relationships, we can investigate co-occurrence within given lexical structures such as ‘a cut(s) b’, ‘a work(s) with b’, etc. This has been explored in machine learning and artificial intelligence research (Turney and Pantel 2010).

What does proximity data tell us?

Proximity data represents a relatively data-driven approach to corpus semantics (and to semantic analysis in Natural Language Processing [NLP], artificial intelligence, data science, and other fields). In linguistics, the use of proximity data in this way is based upon the idea that words occurring together or in similar contexts are likely to share a similar meaning or occupy a similar conceptual field. This is known as a contextual theory of meaning, and in its early stages the theory was developed in particular by J. R. Firth, Michael Halliday, and John Sinclair (cf. Stubbs 1996; Oakey 2009). Sinclair pioneered the application of the theory in lexicography, with the Collins COBUILD Dictionary. That dictionary designed its entries around the most frequent collocational patterns for each dictionary headword, as evidenced by corpus data. In addition to lexicographical applications, proximity data are now used to study lexical semantics; to automatically identify Parts of Speech; to generate computer models of linguistic meaning in NLP and artificial intelligence studies; as well as to engineer text search tools, summarise texts, identify text topics, and even analyse writers’ ‘sentiment’ (cf. Manning and Schuetze 2001, Chapter 5).

But there is a crucial epistemological question that arises here. At its most basic level, co-occurrence data in corpora tell us directly about language use and usage. What is the link between corpus data showing lexical usage, on the one hand, and lexical semantics or conceptual fields, on the other? That is a question that will preoccupy Linguistic DNA as it evolves – and a question we will continue to address on the blog.

Works Cited

Gries, Stefan Th. 2012. Behavioral profiles: A fine-grained and quantitative approach in corpus-based lexical semantics. In Gary Libben, Gonia Jarema and Chris Westbury (eds), Methodological and analytic frontiers in lexical research. Amsterdam: John Benjamins Publishing Company. 57-80.

Hilpert, M. 2012. Diachronic collostructional analysis meets the noun phrase. In T. Nevalainen and E. C. Traugott (eds.), Oxford Handbook of the English Language. Oxford 2012. 233–44.

Manning, Christopher and Hinrich Schuetze. 2001. Foundations of statistical natural language processing. Boston: MIT Press.

Oakey, David. 2009. Fixed collocational patterns in isolexical and isotextual versions of a corpus. In Paul Baker (ed.), Contemporary corpus linguistics. London, Continuum. 140-58.

Stefanowitsch, Anatol & Stefan Th. Gries. 2008. Channel and constructional meaning: A collostructional case study.  In Kristiansen and Dirven (eds.), Cognitive Sociolinguistics: Language variation, cultural models, social systems, 129-152. Berlin: Mouton de Gruyter.

Stubbs, Michael. 1996. Text and corpus analysis. Oxford: Blackwell.

Turney, Peter D. and Patrick Pantel. 2010. From Frequency to Meaning:
Vector Space Models of Semantics. Journal of Artificial Intelligence Research 37, 141-188.