INFORMATION RETRIEVAL
It aims at finding information in lots of documents in an electronic format. The users of these systems make consultations (searches) expressing what contents they want to find. That is why it is necessary that the system processes previously the collection of documents in order to build access structures (indexes) permitting to find quickly. The Internet search engines are the most popular information retrieval systems.
You can distinguish different types of information retrieval systems or, which is the same, different types of search, depending on the format of the documents with which you work, the languages in which they are and the characteristics of the consultation. At Daedalus we distinguish:
Multimedia search. It consists of searches on videos, audios, images and/or texts. In order to do it, complementing the exploitation of metadata associated to this type of contents, you get a group of high and low-level descriptors for each format. Sometimes the voice recognition techniques and, in other situations, characteristics of colour, brightness, etc. are exploited to get transcriptions of the audio contents. In our demo website you can access our search system on videos, DALI (Digital Audio Library Indexing), as well as a tool to find images similar to others, called System of images search based on Content.
Multilingual search. It aims at finding information in a language starting from another language. For example, you can find texts in English starting from consultation phrases in Spanish, or texts in Spanish starting from a consultation in Catalan. If you want to test it, access the demo available in multilingual search, in our demo website. Our product STILUS Trans incorporates the Daedalus technology of translation used by clients like Yell Publicidad and partners like FutureSpace.
Fuzzy search. Sometimes the lists of results are empty because the words used in the search phrases contain spelling mistakes. In these situations it would be useful to find in the documents words that are similar to the search word. This type of search is called fuzzy search. You can test its efficiency in our fuzzy search demo. The STILUS Fuzzy product contains this technology, used by clients like Yell Publicidad or SGAIM.
Daedalus has developed tools in the field of information retrieval among which you can find:
Content-based image retrieval (CBIR): The CBIR systems are based on graphic features of the images such as the colour, the brightness, the texture and the spatial distribution to recognize similar standards in the different images. Through different functions, the CBIR systems extract a great number of features from each image. Our application is based on a CBIR system that has a free code named FIRE - Flexible Image Retrieval Engine, with different adaptations and improvements in the extraction of features of the images and the measures of distance, which optimizes the database used and improves the quality of the results. Our tool makes possible to search in a database of art images, so that you can find paintings as well as their painters or more information by uploading the image you want.
Look4leaks: It is a tool to search for Wikileaks cables according to different criteria (date, origin of the communication or degree of confidentiality) in five languages: Spanish, English, French, Italian and Portuguese. For that purpose, tools of automatic translation, as well as components of named entity recognition, fuzzy search, semantic tagging, etc. have been used.
Don't forget you can access our demo website, Showroom, to know more real examples of application of our technologies.
Do you want to know more about the Daedalus products for information retrieval?
