Annotating events, Time and Place Expressions in Arabic Texts

Proceedings of Recent Advances in Natural Language Processing, pages 25–31, Hissar, Bulgaria, 7-13 September 2013. Annotating Events, Time and Place Expressions in Arabic Texts Abstract We present in this paper an unsupervised approach to recognize events, time and place expressions in Ara- bic texts. Arabic is a resource –scarce language and we don’t easily have at hand annotated corpora, lexi- cons and other needed NLP tools. We show in this work that we can recognize events, time and place expressions in Arabic texts without using a POS an- notated corpus and without lexicon. We use an unsu- pervised segmentation algorithm then a minimalist set of rules allows us to get a partial POS annotation of our corpus. This partially annotated corpus will serve as a basis for the recognition process which imple- ments a set of rules using specific linguistic markers to recognize events, and expressions of time and place. 1 Introduction The considerable development of information and communication technology has fundamental- ly changed the way we access knowledge. To deal with the huge volumes of information, con- stantly increasing, efficient and robust technolo- gies are needed. In this context, named entities (persons, places, organizations, dates ...) are re- quested in order to categorize, index, summarize, this information. A very useful resource for conducting research in the area of NLP is an annotated corpus which can be used as data in the development of algo- rithms and as data in the evaluation of those al- gorithms (Mazur, 2012). However, natural lan- guages are not all equal regarding the availability of such corpora. Arabic is among the resource- scarce languages and the Arabic NLP (ANLP) community still suffers from the lack of free available annotated corpora, electronic lexicons and other needed NLP tools. Moreover, there are no established (theoretical) linguistic studies to rely on, in the field of NER though there is re- cently an increasing interest from the ANLP community. We propose in this work a minimal- ist approach that allows recognition and annota- tion of key expressions in a raw corpus using only formal indices in the texts. This is not an exhaustive annotation of NEs but rather an em- pirical approach to provide a useful ANLP re- source. The rest of the paper is organized as fol- lows: section 2 is a survey of related work, sec- tion 3 describes our minimalist approach to event, time and place expressions recognition in Arabic texts, section 4 reports the results and evaluation of the approach and finally, we end with a conclusion and future work in section 5. 2 Related work In the growing field of Information Extraction (IE), Named Entity Recognition (NER) refers to the recognition and categorization by types of person names, organizations, locations, numerals as well as time/dates. Nadeau and Sekine (2009) provide a pretty large survey of work on NER where we can find a large variety of NER tools for a few widely used languages. There are gen- erally three main approaches to NER. Linguistic …

Content truncated to 500 words