Generation of named entities

Generation of Named Entities Marisa Jiménez Microsoft Research One Microsoft Way Redmond, WA 98052 USA marialj@microsoft.com Abstract In this paper we present an overview of an approach developed at Microsoft Research to generate strings for named entities such as places and dates. This approach uses abstract representations as input. We first provide an overview of our system to identify named entities in text. Next we present our approach to generate these entities from abstract representations, known as “logical forms” in our system. We then focus on the generation of place names in Spanish. We discuss our technique to generate Spanish place names from a logical form where language-specific features, such as word order, or capitalization conventions do not exist. We finally present the details of a study that we carried out to help us make sound linguistic decisions in the generation of place names in Spanish. Keywords Generation, MT, factoids, named entities, logical form 1. Introduction Proper identification of multi-word expressions that refer to proper names, dates, company names, places, and other entities is a need for most natural language processing (NLP) applications. Information retrieval is probably one of the most popular NLP applications to make full use of name identification techniques (Mani et al, 1993; Cowie and Lehnert, 1996; Paik et al, 1993, among others). Named entity identification is also important for Machine Translation (MT) systems. MT systems should have some mechanism for identifying person names such as Mr. John Little. They should also know that Mr. is a title word that should be translated, and that they should avoid translating John Little into Spanish as Juan Pequeño (Wacholder et al, 1997). In this paper we first present an overview of the system developed by the NLP group at Microsoft Research to identify named entities. Our system uses what we call factoid rules to identify multi-word expressions that are productive and not in our current dictionaries. We then describe our approach to the generation of named entities. The factoid generation module makes use of an abstract representation to generate different named entities. This module is still under development. At the moment we have rules that generate dates, units, and places. 2. Identification of Named Entities in the Microsoft NLP System The Microsoft NLP system uses rules to identify multi- word expressions that are productive and not part of our monolingual dictionaries. These expressions are known as factoids and the rules we used to identify them are known as factoid rules. Factoid rules look like simple grammar rules that apply before the grammar rules do. They are written in G, a Microsoft-internal high-level linguistic formalism used as the development tool by our group.1 1 See Heidorn (1972) for details about this linguistic formalism. They can apply recursively, and can also be the input to other factoid rules. Factoid rules were developed to help the analysis component of our system deal with multiple word expressions that were not defined in our dictionaries. The rules can be divided into two groups: those …

Content truncated to 500 words