JPH0498463A - Photograph retrieving device - Google Patents

Abstract

PURPOSE:To take out the language data of similar language and of same meaning by butting photograph data and the data taken out from a photograph attribute storing part, referring to a similar language storing part, calculating similarity and outputting indication so as to display from the photograph data of which a score is large. CONSTITUTION:A dictionary storing part 8 stores every word to represent the contents of a photograph and the accessory information of it (a part of speech classification, a kind, an attribute and so on) and is referred to from an analyzing part 4 at the processing of input language. A photograph attribute storing part 9 stores information to represent features and so on to be the retrieving key of the photograph stored in a photograph storing part 11, that is, an item name, the word, the word attribute. A similar language storing part 10 stores the combination of various retrieving terms and words and a point to represent the similarity of them and is inquired by a retrieving part 5. Thus, retrieving to make the most of the information input of natural language can be realized and the objective data can be retrieved irrespective of the way of the representation of a person to retrieve by using dictionary data and similar language data.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an information retrieval device using a simple natural language interface.

[Conventional technology]

Conventionally, when performing a search process, a plurality of attributes are attached to individual photo data stored in a photo search device, and these are used as search keys for searching. In other words, the attributes of the photo you want to search are input as a search request, and the search request is checked against the attributes of the photos stored in the device to see if the search request is satisfied. Photo data with matching or partially matching attributes are displayed as search results.

[Problem to be solved by the invention]

The conventional photo search device described above retrieves only the data that matches the language of the attribute data in the device when a search request is input in natural language, so even if the data has the same language that means the same thing, it is the same. If the language is not attached to the data, it will be hidden from the search results. The way of expressing things is different for each person, and there are many languages that have similar or similar meanings, but it is not possible to extract such data. When data that matches the search request is found. ,
Generally, there are a large number of these, and data is retrieved in the order in which they are discovered, so the user conducting the search has to check all the search results, which has the disadvantage that the checking process is not efficient.

[Means to solve the problem]

The photo search device of the present invention includes an input unit that accepts natural language, a dictionary storage unit that stores attached information to be added to analyzed words, and a photo attribute storage unit that stores attribute data attached to photo data. A similar word storage unit that stores the degree of similarity between words; and a similar word storage unit that analyzes the input natural language and adds attached information, and compares this data with the data retrieved from the photo attribute storage unit to store the similar word storage unit. The apparatus includes a control section that calculates the degree of similarity by referring to the section and issues an instruction to display photo data with a higher score first, and a display section that displays the photos instructed by the control section.

〔Example〕

Next, the present invention will be explained with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the present invention. In the figure, the photo search device includes an input section 1.

It has a display section 2, a control section 3, a dictionary storage section 8, a photo attribute storage section 9, a similar word storage section 10, and a photo storage section 11.

The input unit 1 accepts natural language information as key information for searching for photos. Information input through the input section 1 is sent to the control section 3.

The control unit 3 analyzes the input natural language, replaces it with a search term, refers to the similar word storage unit 10, calculates the degree of similarity between the above search term and the attribute of the photo attribute storage unit 9, and selects a keyword with a high degree of similarity. An instruction is given to the display section 2 to display the picture first. In order to realize these functions, the control section 3 includes an analysis section 4. Search section 5. Score calculation part 6. A sorting section 7 is provided.

The analysis unit 4 analyzes the natural language sent from the input unit 1 and classifies it into the language expressing the item (item name) and the language expressing the item (word).

Then, referring to the information stored in the dictionary storage section 8, a word attribute 22 and a search term 23 are assigned to each word (see FIG. 2). These pieces of information are sent to the search section 5.

The search unit 5 retrieves the corresponding word from the photo attribute storage unit 9 based on the item name classified by the analysis unit 4 and the word attribute assigned, and uses the item name, word attribute, search term, and word as a search key. The similar term storage unit 10 is searched and a score representing the degree of similarity between the search term and the word is extracted. The extracted score is sent to the score calculation section 6.

The score calculation section 6 multiplies all the scores sent from the search section 5 to calculate a total score.

The sorting section 7 rearranges the photo IDs sent from the score calculation section 6 in descending order of total score, sequentially takes out the photos from the photo storage section 11 according to the order of the IDs, and issues a display instruction to the display section 2.

The display unit 2 displays the photo requested to be displayed by the control unit 3.

The dictionary storage unit 8 stores all words representing the content of the photograph and their associated information (one type for each part of speech, attributes, etc.), and is referred to by the analysis unit 4 when processing the input language.

The photo attribute storage unit 9 stores information stored in the photo storage unit 11 that represents the characteristics, etc., that serve as a search key for the photos, i.e., item name 3.
1. Word 33. Word attributes 32 are stored (see FIG. 3), and these are referred to by the search unit 5 when searching for similar words.

The similar term storage unit 10 stores various combinations of search terms and words and scores representing their degrees of similarity, and the search unit 5 is often referred to.

The photo storage section 11 stores photo data.

Next, the operation of the above-mentioned photo search device will be specifically described for searching for photos of dogs.

First, the dog's characteristics are input in natural language through the input unit 1. For example, assume that these data are input: ``J has a small head, J has round eyes,'' ``J has long hair, and J has floppy ears.'' The input language is sent to the control unit 3.

The sent language is analyzed by the analysis unit 4 and classified into item names and words. For example, the item name is ``head'', and the words are ``small'', ``eyes'', ``sharp'', and ``hair''. "long"
, "mimi" and "tateru" are classified in this way. The dictionary storage unit 8 is searched using each classified word as a key, and the information on the corresponding word, that is, the word attribute 22 and the search term 2, is searched.
Withdraw 3 and give 0 For example, referring to Figure 2, “
The word attribute of the word "small" is "large", and the search term is "small" as is. Furthermore, the word attribute of the word "kuritsutoshiru" is "shaped", and the search term is "round". The data to which word attributes and search terms have been added in this way is sent to the search unit 5.

The search unit 5 searches the photo attribute storage unit 9 using the item names classified by the analysis unit 4 and the word attributes assigned as keys, and retrieves the corresponding photo attributes and words 33. For example, referring to Figure 3, in the photo attribute data of the photo, the item name is "r".
The word with the word attribute "size" is "large", and this word is extracted as data on the photo side.

Next, the similarity storage unit 10 is searched using the search term assigned by the analysis unit 4, the word extracted as data on the photo, the item name, and the word attribute as keys, and scores representing similar words are retrieved. is "head", word attribute "size J,
If the search term is "F small" and the word on the photo side is "large," the similar word storage section 10 is searched using these keys.

First, the item name "head" is used as the first key to search, and then the word attribute "size" is used to search for corresponding data. The relevant data retrieved at this time has a structure as shown in FIG. A combination of the search term ``small j'' and the word ``large'' is found from this data, and the score for that combination is extracted. Applying it to Figure 4, if we look at the search term "small" horizontally and the word "large" vertically, the score for this combination of "small-large" is ro,
Becomes IJ. In this way, the analysis unit 4 extracts scores for all the words analyzed and classified, and all the extracted scores are sent to the score calculation unit 6.

The score calculation unit 6 multiplies all the sent scores together and calculates 1
Calculate the total score for the input data for the photos. All photo data stored in the photo attribute storage section 9 is processed by the search section 5, a total score is calculated, and the total score is sent to the sorting section 7.

The sorting unit 7 rearranges the sent total scores in descending order of scores, and at the same time rearranges the photo IDs. In accordance with the order of the photo IDs, the photos are taken out from the photo storage section 11 and an instruction to display them is given to the display section 2.

The display unit 2 displays the photo for which the instruction has been requested.

〔Effect of the invention〕

As explained above, the present invention realizes a search that takes advantage of natural language information input by using dictionary data and similar word data, and searches for target data regardless of the way the searcher expresses it. I can do it. Furthermore, by displaying photos with the highest degree of similarity between the input natural language and photo attributes, confirmation of multiple search results becomes more efficient.

[Brief explanation of drawings]

Fig. 1 is a configuration diagram showing an embodiment of the present invention, Fig. 2 is an explanatory diagram showing an example of dictionary data stored in the dictionary storage section, and Fig. 3 shows an example of photo attribute data stored in the photo attribute storage section. FIG. 4 is an explanatory diagram showing an example of similar word data stored in the similar word storage section. 1...Input section, 2...Display section, 3...
...Control section, 4...Analysis section, 5...
・Search section, 6...Score calculation section, 7...
・Sort section, 8... Dictionary storage section, 9...
・Photo attribute storage unit, 10...Similar word storage unit, 1
1...Photo accumulation department.

Claims (1)

[Claims] An input unit that accepts natural language, a dictionary storage unit that stores attached information to be added to analyzed words, a photo attribute storage unit that stores attribute data attached to photo data, and stores similarities between words. Analyzes the input natural language and adds additional information to it, compares this data with the data retrieved from the photo attribute storage section, and calculates the degree of similarity by referring to the similar word storage section. a control unit that calculates and instructs to display photo data with a higher score;
A photo search device comprising: a display section that displays a photo instructed by the control section.