Working skill estimating program

• US 8,341,012 B2
• Filed: 07/27/2005
• Issued: 12/25/2012
• Est. Priority Date: 03/17/2005
• Status: Active Grant

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Yamanaka, Hideki, Yanase, Takashi, Namba, Isao
• Primary Examiner(s): Jeanty, Romain
• Assistant Examiner(s): FEENEY, BRETT A
• Application Number:
  US11/189,825
• Publication Number:
  US 20060210052A1
• Time in Patent Office: 2,708 Days
• Field of Search: None
• US Class Current: 705/7.41
• CPC Class Codes: H04M 2201/14 Delay circuits; Timers
• CPC Class Codes: H04M 3/5233 Operator skill based call distribution

Abstract

A working skill estimating program is provided for estimating working skills of each agent in the state of separating a disturbance factor from response records (incidents). A large category classifier is supplied with the incidents within a predetermined time and classifies the incidents according to classification rules, for creating the topic-classified incidents. A small category generator collects the analogous incidents for each task type of the topic-classified incidents and generates a small category. A disturbance distinguisher distinguishes if the incidents belonging to each small category are common response records that appear regularly over time or disturbances whose frequency exhibits a significant time dependency. Then, a skill estimator estimates the working skills of each agent for each task type by using only common incidents determined as the common information.

Claims

1. A computer-readable medium encoded with a working skill estimating program for estimating working skills of an agent, the program, when executed by a computer, causing the computer to function as:

a large category classifying unit to classify response records collected during a predetermined period into large categories according to predetermined classification rules based on topics of inquiries and responses recorded in the response records;

a small category generating unit to generate small categories by further classifying the response records belonging to each of the large categories according to a predetermined definition of similarity between individual response records;

a disturbance distinguishing unit to classify the response records belonging to each of the generated small categories into common response records that are produced constantly over time or disturbances whose frequency exhibits a significant time dependency; and

a skill estimating unit to select one of the large categories, extract the response records belonging to the selected large category while removing therefrom the response records of the small categories that the disturbance distinguishing unit has classified as disturbances, and estimate working skills of the agent in the selected large category, based on at least how many response records are extracted.

2. The computer-readable medium according to claim 1, wherein the small category generating unit uses a hierarchical clustering to collect similar response records, the small category generating unit comprising:

a hierarchy building unit to build a hierarchical structure of clusters by repetitively producing a new upper cluster from two lower clusters having the highest similarity at present which are selected from among the clusters each containing at least one response record, based on the definition of similarity; and

a stop determining unit to determine when to stop the process of building the hierarchical structure by the hierarchical building unit, according to a predetermined stop condition, so as to finalize the hierarchical structure of clusters.

3. The computer-readable medium according to claim 2, wherein:

the hierarchy building unit evaluates the similarity between two clusters, based on a combination of first and second similarities thereof;

the first similarity is defined as an absolute difference between a ratio of the response records created during a first half of the predetermined period and a ratio of the response records created during a second half of the predetermined period; and

the second similarity is similarity in content between every two response records belonging to the two clusters in question.

4. The computer-readable medium according to claim 3, wherein the first similarity is higher as the absolute difference between the ratios is smaller.

5. The computer-readable medium according to claim 3, wherein:

the small category generating unit compares every possible pair of response records each belonging to the two clusters in question, so as to evaluate similarity in terms of words contained therein; and

a higher value of the second similarity is given to the two clusters if a higher minimum similarity is observed in the words contained therein.

6. The computer-readable medium according to claim 2, wherein:

the stop determining unit determines whether the stop condition is met each time a new cluster is created; and

the stop condition specifies either or both of a maximum number of response records that each cluster is allowed to contain and a minimum similarity requirement that every two clusters to be coupled should satisfy.

7. The computer-readable medium according to claim 1, wherein:

the classification rules each give a particular large category name in association with a set of keywords that are supposed to appear in response records of that large category; and

the large category classifying unit retrieves the keywords contained in the response records and, based on the classification rules, allocates to the response records the large category name associated with the retrieved keywords.

8. The computer-readable medium according to claim 7, wherein:

each response record comprises an inquiry content field containing details of an inquiry and a response content field containing details of a response to the inquiry; and

the large category classifying unit retrieves the keywords from either or both of the inquiry content field and the response content field.

9. The computer-readable medium according to claim 1, further causing the computer to function as a disturbance analyzing unit to calculate how much the disturbances have influenced productivity of the agent, by using the response records distinguished as the disturbance information or the common information by the disturbance distinguishing unit.

10. The computer-readable medium according to claim 9, wherein:

the disturbance analyzing unit calculates the influence of the disturbances, based on a relative productivity influenced by the disturbance in a first half of the predetermined period, a relative productivity influenced by the disturbances in a second half of the predetermined period, a ratio of common response records to a whole of the response records in the first half of the predetermined period, and a ratio of common response records to the whole response records in the second half of the predetermined period.

11. The computer-readable medium according to claim 1, wherein the disturbance distinguishing unit classifies the response records by dividing the response records in a given small category into two groups depending on whether the response records are collected in a first half or a second half of the predetermined period, comparing ratios thereof, and classifying the response records in the given small category as common response records when said comparison. finds substantially equal ratios of the two groups, or as disturbances when said comparison finds different ratios of the two groups.

12. A working skill estimating apparatus for estimating working skills of an agent, the apparatus comprising a computer with a storage unit which operates as:

a large category classifying unit to classify response records which are collected during a predetermined period and stored in the storage unit of the computer into large categories according to predetermined classification rules based on topics of inquiries and responses recorded in the response records;

a small category generating unit to generate small categories by further classifying the response records belonging to each of the large categories according to a predetermined definition of similarity between individual response records;

a disturbance distinguishing unit to classify the response records belonging to each of the generated small categories into common response records that are produced constantly over time or disturbances whose frequency exhibits a significant time dependency; and

a skill estimating unit to select one of the large categories, extract the response records belonging to the selected large category while removing therefrom the response records of the small categories that the disturbance distinguishing unit has classified as disturbances, and estimate working skills of the agent in the selected large category, based on at least how many response records are extracted.

13. A working skill estimating method executed by a computer with a storage unit for estimating working skills of an agent, the method comprising:

(a) classifying, by the computer, response records which are collected during a predetermined period and stored in the storage unit into large categories according to predetermined classification rules based on topics of inquiries and responses recorded in the response records;

(b) generating, by the computer, small categories by further classifying the response records belonging to each of the large categories according to a predetermined definition of similarity between individual response records;

(c) classifying, by the computer, the response records belonging to each of the generated small categories into common response records that are produced constantly over time or disturbances whose frequency exhibits a significant time dependency; and

(d) selecting, by the computer, one of the large categories, extracting the response records belonging to the selected large category while removing therefrom the response records of the small categories that said classifying has classified as disturbances, and estimating working skills of the agent in the selected large category, based on at least how many response records are extracted.