Research On Food Complaint Document Classification Based-on Topic

The food is the material base that human survived and developed. It is important forevery people’s life. In recent years, vicious food security accident happened frequently, andthat attracted people’s high attention. The customer is the direct experience of food, and theirresponse is very important to discover food hazard, so their response for the food plays asignificant role in finding fake and quality food. With the rapid development of Internet andthe improved food security monitoring system, the customer already can join in themanagement of food security by complaining on the Internet. However, the issued foodcomplaint documents by customer on the Internet are out of order, which are not effectivelyorganized. And different information users have different require for complaint information. Itis a burning question that how to organize and manage these complaint text effectively, andclassify them according to the topic, which can help people find the information what theywant rapidly and accurately.Text classification technology is seen as the key method to deal with and organizenumerous documents. It can help people accurately find the information what they want.Traditional text classification methods are essentially used for dealing with the commondocuments (i.e., long documents), but food complaint documents in the real world mostly areshort texts. Because the length of complaint documents is short, which have the inherentshortcomings that features are not enough and not clear, the first question we should solve isfinding an effective semantic expanding method to supplement and intensify the features.The paper proposes a text classification schema which based on the created domainontology and the general ontology HowNet, that we introduce ontology as backgroundknowledge to expand the semantic of complaint documents and supplement effectiveinformation, thus improving the effect of text classification. The difficulty is how to combinecreated food domain ontology and existed general ontology HowNet, and then we used themto expand terms for each topic, thus we can achieve semantic expanding for food complaintdocuments. Finally, we make use of the obtained expanding topic term vectors to calculate thesimilarity with the unlabeled documents, thus we realize the automatic topic classification forfood compliant documents.