SUBJECT: Re : &NAME Introduction &NAME - &NAME did n't appear to be cc' &NAME on this email . Good stuff below . Naturally , the text below would be broken up into different part of this web site ( or least divided into headings ) . &NAME * Definitely mention which markets that ilexir ' talks to ' . &NAME ' which currently plays a critical role in the information extraction which has proven vital to the success of the [ &NAME ] market , which rapidly escalating in the importance in the knowledge-based economy . ' * &NAME of the Product ( suggest easy on the consultancy . Naturally , there will be implementation , best practice assessment etc - which we know will be the cash cow in the short term etc. Never mention costs , daily rates should only be discounted under extreme conditions and never mentioned up front ) * Simple Example ( possibly linked to a &NAME &NAME ) * Detailed Textual Analysis * Which leads to : What are the advantages ? Or why should you consider ilexir instead of the competition ? ( Best of breed , emphasise University heritage , years of rigorous testing , proven . &NAME I ca n't tell if &NUM is good or exceptionally good . ) Best , &NAME . &NAME / &NAME &NAME , thanks for making the introduction . As promised , I 'm sending you a bit more draft text that addresses in less technical terms what we are about and what makes the technology unique . When finished this , will be part of the ilexir website . Comments very welcome . I 'm also including some urls pointing to the existing non-commercial licence and quick demo for researchers of the system . &NAME , if you think there is a potential connection / collaboration here , then please do get back to me best , &NAME &NAME ( &NAME in Computational Linguistics ) &ORG University of &NAME &NAME &NAME Ave &NAME &NAME &NAME , &NAME Email : &EMAIL Web page : &WEBSITE &NAME : &NUM &NUM &NAME : &NUM &NUM &WEBSITE &NAME &NAME provides commercial licensing , support and consultancy for the robust accurate statistical parsing ( &NAME ) toolkit . The toolkit allows rapid development of advanced text management applications such as text classification , summarisation , mining and querying . It is uniquely capable of supporting the next generation of such systems , which go beyond pure term indexing or domain-specific entity extraction . The &NAME toolkit is a package of text processing modules which annotate text with linguistic information concerning the entities referred to , properties predicated of these relations , and relations which hold between them . For instance , the following text sentence-from the transcript of President &NAME 's speech at the &NAME Museum Center on 7th October , &NUM -- is analysed into sets of entities and relations between them some of which are shown underneath : Members of the &NAME of both political parties , and members of &ORG &ORG , agree that &NAME &NAME is a threat to peace and must disarm . Entities : [ Member+s of the &NAME of both political parties ] [ Member+s ] [ the &NAME ] [ both political parties ] [ political parties ] [ parties ] [ &NAME &NAME ] [ &NAME ] [ &NAME ] [ &ORG &ORG ] [ &ORG ] [ the &ORG ] [ the Council ] Relations : [ the &NAME ] agree+s [ [ &NAME ] be / is [ a threat ] ] [ the Council ] agree+s [ [ &NAME ] be / is [ a threat ] ] [ the &NAME ] agree+s [ [ &NAME ] [ must disarm ] ] [ the Council ] agree+s [ [ &NAME ] [ must disarm ] ] The actual output of the system further decomposes and classifies these relations and there are other output formats ( href ) Term-based systems represent text documents as ' bags of words ' ( i.e. sets of characters between white spaces ) but will not directly recover the entities referred to in a document , such as ' &NAME &NAME ' , ' political parties ' , ' &ORG &ORG ' or ' members of &NAME of both political parties ' . Representing entities rather than terms supports more powerful and precise information management because it provides a better basis for cross indexing variants both within and across documents , such as ' the &NAME &NAME Council ' or ' the &ORG ' , and thus also a better way of classifying , summarising or querying document content via salient entities . The &NAME system recovers noun phrases referring to entities and classifies them in a domain-independent way into names ( subdivided into places , people , organisations ) , numbers ( including ranges , dates , etc ) , measure phrases ( ounce , year , etc ) , temporal expressions ( days , weeks , months ) , directions ( north , south , etc ) , and so forth . Relational representation of entities provides a natural and effective mechanism for cross-indexing alternate descriptions of entities via subsets of relations ( href ) . Extant systems for extracting entities from text , based on deterministic finite-state pattern matching , rely heavily on manually-supplied domain-specific lexical information from the outset . The relational information recovered by the &NAME system can be used to index ' who did what to whom ' and supports precise querying about the properties predicated of specific entities across a document collection . A relational pattern like ' [ the &NAME &NAME Council ] agree [ ? ] ' ( ' What does the &NAME &NAME Council agree ? ' ) can be matched intelligently against document annotations yielding precise , specific and small amounts of focussed information . This approach to information retrieval deals naturally with non-contiguous text sequences mapping to the same underlying entity or relation ( [ the &ORG ] [ the &NAME &NAME Council ] ) as well as other well known problems with pure term or entity based systems , such as negation ( [ the Council ] [ not agree ] [ [ &NAME ] [ must disarm ] ] $ neq $ [ the Council ] agree [ [ &NAME ] [ must disarm ] ] ) . The &NAME toolkit achieves about &NUM accuracy recovering entities and relations from arbitrary English texts , without making use of domain-specific lexical information . It has been used to date to analyse about &NUM million words of English . A recent paper from a refereed conference proceedings gives more details about each module of the toolkit and compares performance to related systems . &NAME and performance for specific applications can be considerably improved by semi-automatic customisation on domain and / or application specific text . The toolkit is fully compatible with existing &NAME standards for text &NAME up and can be used to leverage web sites which will conform with emerging standards for the semantic web ( &NAME / &NAME href ) . A non-exclusive commercial licence to use the &NAME toolkit , which includes &NUM hours support and consultancy to assist with customising and integrating the toolkit into the client 's application , costs $ pounds $ &NUM , &NUM . The toolkit is available free for non-commercial use including prototyping and proof-of-concept work in industry . Support for this and / or general consultancy on &NAME / IR applications and systems is available from &NAME Ltd at a rate of $ pounds $ &NUM / hour . The cofounders of &NAME are &NAME &NAME and &NAME &NAME . The &NAME toolkit is the product of collaborative research at the University of &NAME , &NAME Laboratory and University of &NAME , &ORG . &NAME &NAME has sole rights to commercial exploitation of the toolkit and works closely with these academic groups to continuously enhance and improve the performance of the toolkit .