The Potential of Automatic Word Comparison for Historical Linguistics

Proceedings of the 15th Conference of the European Chapter of The Association for Computational Linguistics: Volume 1

Sofroniev³

2017

Self Cite

Most current approaches in phylogenetic linguistics require as input multilingual word lists partitioned into sets of etymologically related words (cognates). Cognate identification is so far done manually by experts, which is time consuming and as of yet only available for a small number of well-studied language families. Automatizing this step will greatly expand the empirical scope of phylogenetic methods in linguistics, as raw wordlists (in phonetic transcription) are much easier to obtain than wordlists in which cognate words have been fully identified and annotated, even for under-studied languages. A couple of different methods have been proposed in the past, but they are either disappointing regarding their performance or not applicable to larger datasets. Here we present a new approach that uses support vector machines to unify different state-of-the-art methods for phonetic alignment and cognate detection within a single framework. Training and evaluating these method on a typologically broad collection of gold-standard data shows it to be superior to the existing state of the art.

Section: Methodsmentioning

confidence: 99%

Using support vector machines and state-of-the-art algorithms for phonetic alignment to identify cognates in multi-lingual wordlists

Jäger

Proceedings of the 15th Conference of the European Chapter of The Association for Computational Linguistics: Volume 1

Sofroniev³

2017

Self Cite

“…Unfortunately, available tools are not up to the task. Computational methods for automatic cognate detection, which could be used to pre-parse the data for the linguists, usually assume that words are morphologically simple (Steiner et al 2011;List et al 2017) and automatic partial cognate detection is still in its infancy .…”

Section: Partial Cognate Annotationmentioning

confidence: 99%

Challenges of annotation and analysis in computer-assisted language comparison: A case study on Burmish languages

Hill

Yearbook of the Poznan Linguistic Meeting

2017

Self Cite

The use of computational methods in comparative linguistics is growing in popularity. The increasing deployment of such methods draws into focus those areas in which they remain inadequate as well as those areas where classical approaches to language comparison are untransparent and inconsistent. In this paper we illustrate specific challenges which both computational and classical approaches encounter when studying South-East Asian languages. With the help of data from the Burmish language family we point to the challenges resulting from missing annotation standards and insufficient methods for analysis and we illustrate how to tackle these problems within a computer-assisted framework in which computational approaches are used to pre-analyse the data while linguists attend to the detailed analyses.

“…Markov Clustering is very popular in biology and was shown to outperform the popular Affinity Propagation algorithm (Frey and Dueck 2007) in the task of homolog detection in biology (Vlasblom and Wodak 2009). As a third method, we follow List et al (2016b) in testing Infomap (Rosvall and Bergstrom 2008), a method that was originally designed to detect communities in complex networks. Communities are groups that share more links with each other than outside the group (Newman and Girvan 2004).…”

Section: Network Partitioningmentioning

confidence: 99%

“…Infomap uses random walks to find the best partition of a network into communities. Infomap is not a classical partitioning algorithm, and we do not know of any studies which tested its suitability for the task of homolog detection in evolutionary biology, but according to List et al (2016b), Infomap shows a better performance than UPGMA in automatic cognate detection.…”

Section: Network Partitioningmentioning

confidence: 99%

Using Sequence Similarity Networks to Identify Partial Cognates in Multilingual Wordlists

Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)

Lopez

Bapteste

2016

Self Cite

Increasing amounts of digital data in historical linguistics necessitate the development of automatic methods for the detection of cognate words across languages. Recently developed methods work well on language families with moderate time depths, but they are not capable of identifying cognate morphemes in words which are only partially related. Partial cognacy, however, is a frequently recurring phenomenon, especially in language families with productive derivational morphology. This paper presents a pilot approach for partial cognate detection in which networks are used to represent similarities between word parts and cognate morphemes are identified with help of state-of-theart algorithms for network partitioning. The approach is tested on a newly created benchmark dataset with data from three sub-branches of Sino-Tibetan and yields very promising results, outperforming all algorithms which are not sensible to partial cognacy.