Language Independent Named Entity Recognition Combining Morphological and Contextual Evidence

Language Independent Named Entity Recognition Combining Morphological and Contextual Evidence Silviu Cucerzan and David Yarowsky Department of Computer Science Center for Language and Speech Processing Johns Hopkins University Baltimore, Maryland, 21218 {silviu,yarowsky}@cs.jhu.edu Abstract Identifying and classifying personal, geographic, in- stitutional or other names in a text is an impor- tant task for numerous applications. This paper de- scribes and evaluates a language-independent boot- strapping algorithm based on iterative learning and re-estimation of contextual and mOrphological pat- terns captured in hierarchically smoothed trie mod- els. The algorithm learns from unannotated text and achieves competitive performance when trained on a very short labelled name list with no other required language-specific information, tokenizers or tools. 1 Introduction The ability to determine the named entities in a text has been established as an important task for several natural language processing areas, including information retrieval, machine translation, informa- tion extraction and language understanding. For the 1995 Message Understanding Conference (MUC-6), a separate named entity recognition task was devel- oped and the best systems achieved impressive accu- racy (with an F-measure approaching 95%). What should be underlined here is that these systems were trained for a specific domain and a particular lan- gnage (English), typically making use of hand-coded rules, taggers, parsers and semantic lexicons. In- deed, most named entity recognizers that have been published either use tagged text, perform syntactical and morphological analysis or use semantic informa- tion for contextual clues. Even the systems that do not make use of extensive knowledge about a partic- ular language, such as Nominator (Choi et al., 1997), still typically use large data files containing lists of names, exceptions, personal and organizational iden- tifiers. Our aim has been to build a maximally langnage- independent system for both named-entity identi- fication and classification, using minimal informa- tion about the source language. The applicability of AI-style algorithms and supervised methods is lim- ited in the multilingual case because of the cost of knowledge databases and manually annotated cor- pora. Therefore, a much more suitable approach is to consider an EM-style bootstrapping algorithm. In terms of world knowledge, the simplest and most rel- evant resource for this task is a database of known names. For each entity class to be recognized and tagged, it is assumed that the user can provide a short list (order of one hundred) of unambiguous examples (seeds). Of course the more examples pro- vided, the better the results, but what we try to prove is that even with minimal knowledge good re- sults can be achieved. Additionally some basic par- ticularities of the language should be known: cap- italization (if it exists and is relevant - some lan- guages do not make use of capitalization; in others, such as German, the capitalization is not of great help), allowable word separators (if they exist), and a few frequent exceptions (like the pronoun "/" in English). Although such information can be utilised if present, it is not required, and no other assump- tions are made in the general model. …

Content truncated to 500 words