Text Categorization

Interactive Tools

The interactive tools are not longer support due to the limited reousrce! USer should ignore information on this page.

From the current page, you can link to several interactive tools based on the JDI methodology:

JDI Text Input will connect you to a Web Tools page with the JDI link on that page selected by default. You will be able to perform Journal Descriptor Indexing by entering into the Input box the text to be indexed. After Go is selected, this text will be categorized by a set of Journal Descriptors (JDs), which are terms from NLM's MeSH Vocabulary File used for manually indexing NLM serials according to discipline. It is possible to specify MeSH terms as input text by preceding each MeSH term with a vertical bar ( | ); when combining natural language words and MeSH terms, the words must precede the terms.

STI Text Input will connect you to a Web Tools page with the STI link on that page selected by default. You will be able to perform Semantic Type Indexing by entering into the Input box the text to be indexed. After Go is selected, the text will be categorized by a set of Semantic Types (STs), which are terms in the Semantic Network in NLM's Unified Medical Language System (UMLS).

JDI PMID Input will connect you to a TCAT (Text Categorization Application Tools) page with the PMID link on that page selected by default. Using this tool, you will be able to enter PubMed IDs (PMIDs) for Journal Descriptor Indexing of the PubMed citations corresponding to the PMIDs as retrieved from PubMed.

STI PMID Input will connect you to a TCAT (Text Categorization Application Tools) page with the PMID link on that page selected by default. Using this tool, you will be able to enter PubMed IDs (PMIDs) for Semantic Type Indexing of the PubMed citations corresponding to the PMIDs as retrieved from PubMed.

JDI MEDLINE Input will connect you to a TCAT (Text Categorization Application Tools) page with the MEDLINE link on that page selected by default. Using this tool, you will be able to use the Import button to select a text file of PubMed citations in MEDLINE format. The MLT Tags menu is used for selecting which part of the citation to index (Title, Abstract, Title & Abstract, MeSH Terms, or Title & Abstract & MeSH Terms). Using this tool, you will be able to perform Journal Descriptor Indexing on the citations in the imported file.

STI MEDLINE Input will connect you to a TCAT (Text Categorization Application Tools) page with the MEDLINE link on that page selected by default. Using this tool, you will be able to use the Import button to select a text file of PubMed citations in MEDLINE format. The MLT Tags menu is used for selecting which part of the citation to index (Title, Abstract, or Title & Abstract). Using this tool, you will be able to perform Semantic Type Indexing on the citations in the imported file.

For all tools, the output is a ranked list, with scores between 0-1, of either JDs (using JDI tools) or STs (using STI tools) that categorize the input text. The output for all tools where input contains natural language words contains two lists of independent results based on different methodologies. The "word frequency" based result uses a normalized word count of words in a training set; the "document count" based result uses the count of training set documents in which words appear. It has not bee determined which methodology is superior.

After selecting a tool above, you can navigate among these tools by selecting Web Tools and TCAT links in the upper right corner. Once in Web Tools, you can select JDI (the same as JDI Text Input above) or STI (the same as STI Text Input above). Once in TCAT, you can select JDI or STI. Within JDI and STI, you can select Inputs, and then PMID (the same as JDI PMID Input or STI PMID Input, respectively, as described above) or MEDLINE (the same as JDI MEDLINE Inputb> or STI MEDLINE Input, respectively, as described above); you can also select Text. Text under TCAT/JDI has the same functionality as JDI Text Input as described above; Text under TCAT/STI has the same functionality as STI Text Input as described above.

JDI and STI may also be accessed by going to the Text Categorization Home Page and selecting Web Tools or TCAT. In the resulting two pages, you can select JDI or STI, and, within TCAT under JDI or STI, you can select PMID or MEDLINE or Text. The default selection under Web Tools is JDI; the default selection under TCAT/JDI is Text. Text under TCAT/JDI has the same functionality as JDI under Web Tools, and Text under TCAT/STI has the same functionality as STI under Web Tools. PMID and MEDLINE input are available only under TCAT. The basic system for JDI/STI indexing of text input is Web Tools, while PMID and MEDLINE input are available only under TCAT.

Under Web Tools, there is an MLT (MEDLINE Tokenizer) selection, where you can enter a PMID and select tags corresponding to PubMed fields in the citation as retrieved from PubMed. The result will be the text corresponding to the selected tags. One of the uses of the tokenizer is to copy and paste words from the result into a text input window for JDI Text Input or STI Text Input.

On pages permitting STI, you will see STRI (Semantic Type Real-Time Indexing) which performs Semantic Type Indexing using a different methodology from STI.