PatentsView Government Interest Results Analysis

1. Extraction Process | 2. Quality Assurance | 3. Results Analysis

PatentsView Government Interest Statement Summary Statistics

The tables below contain descriptive statistics that give information about the final extracted list of organizations for patents granted between 1976 and 2019. Table 1 highlights the number of organizations that were identified by patent number. Table 2 lists the total number of patents associated with the 10 most frequently occurring organizations. “U.S. Government” is an organization only if the raw government interest statement matches the string “govern*” (with * representing a wildcard that can be replaced with anything) and a specific agency was not identified.

Table 1. Number and Percentage Distribution of Organizations Identified per Patent (Patents Granted Between 1976 and 2019)

Table 2. Ten Most Frequently Occurring Organizations (Patents Granted Between 1976 and 2019)

Table 3 represents a sample of the final output where the “raw organization” column contains the output of the Named Entity Resolution (NER) algorithm and the “cleansed organization” column contains the output of the string-matching algorithm. This is to enable comparison of the output from each algorithm. In most cases, NER produced accurate output. However, in some cases NER either produced an incomplete name of a government organization or identified a non-federal organization. For example, for patent number 5320962, NER produced “Blood Institute” as the organization name instead of National Heart, Lung, and Blood Institute. Similarly, for patent number 5583736, NER identified both the Department of Energy and the American Telephone and Telegraph Company, where the latter is a private company that does not meet the criteria for inclusion in the final output of government organizations.

Table 3. Sample Output of Correctly Identified Cleansed Organizations

Table 4. Sample Output, Incorrectly Identified Cleansed Organizations  Table 4 shows incorrect output in terms of cleansed organizations. In the first record, the list of cleansed organizations should read “Department of Health, Education, and Welfare,” which was disassembled into DHHS and the Department of Education. The NER for patent 7910574 (incorrectly) identified DLW as an organization but not NIH or NIMH. Finally, the last record is not an Army-related patent; this represents an apparent false positive resulting from the approach to organization names.

Contract and Award Number Output

Descriptive statistics were also produced for contract and award numbers. Table 5 summarizes the number of contracts or awards that were identified by patent number, while table 6 shows records with incorrectly identified contract or award numbers. In these cases, a contract number was missed due to its abnormally short length (and the fact that it contains a space), due to a public law being conflated for a contract or award number, and due to bad input data. The third record in Table 6 shows an example of correctly parsed contract and award data.

Table 5. Number of Contracts/Awards Identified per Patent  (Patents Granted Between 1976 and 2019)

Table 6. Examples of Incorrect Contract Output