System and method of detecting common patterns within unstructured data elements retrieved from big data sources

Abstract

A method and system for detection of common patterns within unstructured data elements. The method includes searching a plurality of unstructured data elements extracted from big data sources to identify a plurality of patches; extracting the identified plurality of patches; generating, by a signature generator system, at least one signature for each patch; identifying common patterns among the at least one generated signature; clustering the signatures having the identified common patterns; and correlating the generated clusters to identify associations between the respective identified common patterns.

Description

TECHNICAL FIELD

The present invention relates generally to the analysis of unstructured data, and more specifically to a system for identifying common patterns within unstructured data elements.

BACKGROUND

With the abundance of unstructured data made available through various means in general and the Internet in particular, there is also a need to provide effective ways of analyzing such data. Unstructured data analysis is a challenging task, as it requires processing of big-data. Big data typically refers to a collection of data sets that are large and complex and cannot be analyzed using on-hand database management tools or traditional data processing applications.

Several prior art solutions can be used to search through big data sources. As a result of the search, relevant data elements may be extracted from such big data sources. However, a problem may occur while trying to search for additional data that may be useful, for example, data containing similar characteristics to the characteristics of the extracted data. Typically, the complexity of the data while analyzing the characteristics of big data, leads to inefficient identification of common patterns. Furthermore, the search as known today may be inefficient because of lack of correlation between data elements extracted from big data sources.

It would be therefore advantageous to provide an efficient solution to analyze big data. It would be further advantageous if such solution would enable correlating between common patterns while analyzing the big data.

SUMMARY

Certain embodiments disclosed herein include a method for detection of common patterns within unstructured data elements. The method includes searching a plurality of unstructured data elements extracted from big data sources to identify a plurality of patches; extracting the identified plurality of patches; generating, by a signature generator system, at least one signature for each patch, wherein the signature generator system includes a plurality of computational cores configured to receive the plurality of patches, each computational core of the plurality of computational cores having properties that are at least partly statistically independent of other of the computational cores, wherein the properties are set independently of each other core; identifying common patterns among the at least one generated signature; clustering the signatures having the identified common patterns; and correlating the generated clusters to identify associations between the respective identified common patterns.

Certain embodiments disclosed herein also include a system for analyzing unstructured data. The system includes a network interface for allowing connectivity to a plurality of big data sources; a processing unit; and a memory connected to the processing unit, the memory containing instructions that, when executed by the processing unit, configure the system to: search a plurality of unstructured data elements extracted from the big data sources to identify a plurality of patches; extract the identified plurality of patches; generate, by a signature generator system, at least one signature for each patch, wherein the signature generator system includes a plurality of computational cores configured to receive the plurality of patches, each computational core of the plurality of computational cores having properties that are at least partly statistically independent of other of the computational cores, wherein the properties are set independently of each other core; identify common patterns among the at least one generated signature; cluster the signatures having the identified common patterns; and correlate the generated clusters to identify associations between the respective identified common patterns.

BRIEF DESCRIPTION OF THE DRAWINGS

The subject matter that is regarded as the invention is particularly pointed out and distinctly claimed in the claims at the conclusion of the specification. The foregoing and other objects, features, and advantages of the invention will be apparent from the following detailed description taken in conjunction with the accompanying drawings.

FIG. 1 is a schematic block diagram of a network system for analyzing unstructured data according to one embodiment.

FIG. 2 is a flowchart describing the identification of common patterns within unstructured data elements retrieved from big data sources according to one embodiment.

FIG. 3 is a block diagram depicting the basic flow of information in the signature generator system.

FIG. 4 is a diagram showing the flow of patches generation, response vector generation, and signature generation in a large-scale speech-to-text system.

DETAILED DESCRIPTION

It is important to note that the embodiments disclosed herein are only examples of the many advantageous uses of the innovative teachings herein. In general, statements made in the specification of the present application do not necessarily limit any of the various claimed inventions. Moreover, some statements may apply to some inventive features but not to others. In general, unless otherwise indicated, singular elements may be in plural and vice versa with no loss of generality. In the drawings, like numerals refer to like parts through several views.

Certain exemplary embodiments disclosed herein allow determining a correlation between unstructured data elements extracted from big data sources comprising unstructured data. The correlation refers to any of a broad class of statistical relationships involving at least two sets of data. The embodiments disclosed herein determine the nature of the relationship between the unstructured data elements. One or more common patterns are identified within the unstructured data elements respective of a signature analysis. Unstructured data refers to pieces of data that do not have a predefined structure and are usually not organized in a consistent and predictable manner.

Typically, the data tends to be recorded in a free text form with little or no metadata codified into fields. Unstructured data may be, for example, a multimedia content, a book, a document, metadata, health records, audio, video, analog data, files, unstructured text, web page, a combination thereof, a portion thereof, and so on. The method described herein determines an association of the unstructured data elements in space and/or time respective of the big data sources.

FIG. 1 shows an exemplary and non-limiting schematic diagram of a network system 100 utilized for describing the various embodiments for analyzing multimedia content and determines common patterns of unstructured data elements extracted from big data sources. A network 110 is used to communicate between different parts of the system 100. The network 110 may be the Internet, the world-wide-web (WWW), a local area network (LAN), a wide area network (WAN), a metro area network (MAN), and other networks for enabling communication between the elements of the system 100.

A big data analyzer, for example, the server 130 is connected to the network 110. The server 130 is configured to correlate between unstructured data elements extracted from big data sources comprising unstructured data as described in detail below. The server 130 typically comprises a processing unit, such as a processor (not shown) that is communicatively connected to a memory. The memory contains instructions that are executed by the processor. The server 130 also includes an interface (not shown) to the network 110.

In one embodiment, a database such as, a data warehouse 150 is connected to the server 130 (either directly or through the network 110). The server 130 is configured to store in the data warehouse 150 information identified and/or generated by the server 130 for further use. Such information may include, signatures generated for the unstructured data elements, common patterns identified between the unstructured data elements, common concepts identified between the common patterns, and so on, as described in greater detail with respect of FIG. 2.

Further connected to the network 110 are a plurality of big data sources 120-1 through 120-n, each of which may contain, store, or generate unstructured data. The big data sources 120-1 through 120-n are accessible by the server 130 through the network 110. The system 100 also includes a signature generator system (SGS) 140. In one embodiment, the SGS 140 is connected to the server 130. The server 130 is configured to receive and serve the unstructured data elements. Moreover, the server 130 is configured to cause the SGS 140 to generate the signatures respective of the unstructured data elements. Each signature is generated for each element of the unstructured data.

The SGS 140 typically comprises a processing unit and a memory maintaining executable instructions. Such instructions may be executed by the processor. The process for generating the signatures for the unstructured data elements, is explained in more detail herein below with respect to FIGS. 3 and 4.

According to the various exemplary embodiments, the unstructured data may include multimedia content elements. A multimedia content element may be, for example, an image, a graphic, a video stream, a video clip, an audio stream, an audio clip, a video frame, a photograph, and an image of signals (e.g., spectrograms, phasograms, scalograms, etc.), and/or combinations thereof and portions thereof.

The server 130 is configured to extract unstructured data elements from at least one big data source 120 and provide such elements to the SGS 140. The SGS 140 is further configured to generate at least one signature for each extracted unstructured data element. The generated signature(s) may be robust to noise and distortion as discussed below with respect to FIGS. 3 and 4. Then, using the generated signature(s), the server 130 is configured to search for common patterns through the signatures. Upon identification of one or more common patterns through the signatures, the server 130 is configured to determine the correlation between the unstructured data elements and the associations thereof.

In one embodiment, the signatures generated for more than one unstructured data element are clustered. The clustered signatures are used to search for a common concept. The concept is a collection of signatures representing elements of the unstructured data and metadata describing the concept. As a non-limiting example, a ‘Superman concept’ is a signature reduced cluster of signatures describing elements (such as multimedia elements) related to, e.g., a Superman cartoon: a set of metadata representing proving textual representation of the Superman concept. Techniques for generating concepts and concept structures are also described in the above-referenced U.S. Pat. No. 8,266,185 to Raichelgauz et al., which is assigned to common assignee, and is incorporated hereby by reference for all that it contains.

FIG. 2 depicts an exemplary and non-limiting flowchart 200 describing a method of detection of common patterns within unstructured data elements retrieved from big data sources according to one embodiment. In an embodiment, the process discussed with reference to FIG. 2 is performed by the server 130 when the signatures are generated by means of the SGG 140.

In S210, unstructured data is received or retrieved from a plurality of big data sources, e.g. sources 120-1, 120-n. The sources may be classified by the data contained therein. For example, big data sources 120-1, 120-n may be related to all images found on the Internet, diagnostic information of a large group of patients, sales information of a large group of retail stores, and so on. As noted above, the unstructured data may be, for example, a multimedia content, a book, a document, metadata, health records, audio, video, analog data, files, unstructured text, web pages, a combination thereof, a portion thereof, and so on. In one embodiment, the processing is separately performed on the unstructured data retrieved/received from big data sources having the same classification. For example, all the images in the Internet would be processed separately from e.g. medical records.

In S220, unstructured data elements are extracted from the data collected from the big data sources, e.g. sources 120-1, 120-n. In one embodiment, the extracted elements are of specific interest, or otherwise of higher interest than other elements comprised in the collected unstructured data. As an example, a product's attributes and the sales volume of the product may be of interest. As another example, a disease type and its characteristics may be of higher interest than the patients' names. In addition, certain keywords may be of specific interest. As yet another example, portions of a multimedia element (e.g., a picture) having entropy level over a predefined threshold may be of more interest than other portions. In one non-limiting embodiment, S220 includes searching patterns or patches in the unstructured data, and extracting such identified patterns or patches. Typically, a patch of an image is defined by, for example, its size, scale, location and orientation. A video/audio patch may be a 1%) of the total length of video/audio clip.

In S230, at least one signature for each extracted unstructured data element is generated. In one embodiment, each of the at least one signature is robust to noise and/or distortion and is generated by the SGS 140 as described below. In S240, it is checked if the number of extracted unstructured data elements are above a predefined threshold, i.e., if there is sufficient information for the processing, and if so, execution continues with S250; otherwise, execution continues with S220.

In S250, the generated signatures are analyzed to identify common patterns among the generated signatures. In one embodiment, a process of inter-matching is performed on the generated signatures. In an exemplary embodiment, this process includes matching signatures of all the extracted elements to each other. Each match of two signatures is assigned with a matching score being compared to a preconfigured threshold. When the matching score exceeds the preconfigured threshold, the two signatures are determined to have common pattern.

In S260, the signatures determined to have common patterns are clustered. In an embodiment, the clustering of the signatures is performed as discussed in detail in the U.S. Pat. No. 8,386,400, entitled “Unsupervised Clustering of Multimedia Data Using a Large-Scale Matching System,” filed on Jul. 22, 2009, assigned to common assignee, and which is hereby incorporated for all that it contains. It should be noted that S250 and S260 can result in a plurality of different clusters. As noted above, a cluster may include a textual metadata.

In S270, a correlation between the created clusters is performed to detect association between common patterns identified in the respective clusters. For example, if the common pattern of cluster A is ‘red roses’ and the common pattern of cluster B is ‘Valentine's Day’, then the correlation would detect any association and/or relationship between ‘red roses’ and ‘Valentine's Day’. In one embodiment, a preconfigured threshold level is used to determine if there is an association between at least two clusters of the created clusters. This preconfigured threshold defines at least a number of signatures to be found in two or more correlated-clusters in order to determine that there is an association between the clusters. The correlation can be performed between each two different clusters of the created clusters.

In S280, a concept structure (or concept) is generated by re-clustering two or more clusters determined to have some sort of association between their patterns. In one embodiment, S280 may also include a process for reduction of the number of signatures in the re-clustered cluster and adding of the respective metadata to the reduced clusters to form a concept structure. In S280, the generated signatures, the concept, the clusters, and the identified common patterns may be stored in a data warehouse 150 for further use. In S290, it is checked whether additional unstructured data is received or retrieved, and if so, execution continues with S210; otherwise, execution terminates.

As a non-limiting example, several sales reports of worldwide retail chain stores are received by the server 130. The reports are analyzed and signatures are generated respective of each element within the reports. An element within the reports may be, for example, a certain product, or a certain product together with the quantity sold. Respective of the generated signatures, common patterns are identified, and then clustered as described above. For example, a first common pattern of a first cluster of signatures indicates that every certain date a significant amount of products which are packed in red packages are being sold. A second common pattern of second cluster of signatures indicates that an extensive amount of jewelry is sold in February. A third common pattern of a third cluster of signatures indicates an increase in sales of alcoholic beverages on the eve of February 14. Upon correlating between the plurality of clusters, consumption habits can be determined with respect to Valentine's Day. By analyzing the common patterns of the sales report, the server 130 enables determination of a common concept related to the plurality of common patterns. In this case, the common concept may be “during Valentine's Day people tend to spend more money.”

FIGS. 3 and 4 illustrate the generation of signatures for the multimedia content elements by the SGS 140 according to one embodiment. An exemplary high-level description of the process for large scale matching is depicted in FIG. 3. In this example, the matching is for a video content.

Video content segments 2 from a Master database (DB) 6 and a Target DB 1 are processed in parallel by a large number of independent computational Cores 3 that constitute an architecture for generating the Signatures (hereinafter the “Architecture”). Further details on the computational Cores generation are provided below. The independent Cores 3 generate a database of Robust Signatures and Signatures 4 for Target content-segments 5 and a database of Robust Signatures and Signatures 7 for Master content-segments 8. An exemplary and non-limiting process of signature generation for an audio component is shown in detail in FIG. 4. Finally, Target Robust Signatures and/or Signatures are effectively matched, by a matching algorithm 9, to Master Robust Signatures and/or Signatures database to find all matches between the two databases.

To demonstrate an example of signature generation process, it is assumed, merely for the sake of simplicity and without limitation on the generality of the disclosed embodiments, that the signatures are based on a single frame, leading to certain simplification of the computational cores generation. The Matching System is extensible for signatures generation capturing the dynamics in-between the frames.

The Signatures' generation process will now be described with reference to FIG. 4. The first step in the process of signatures generation from a given speech-segment is to breakdown the speech-segment to K patches 14 of random length P and random position within the speech segment 12. The breakdown is performed by the patch generator component 21. The value of the number of patches K, random length P and random position parameters is determined based on optimization, considering the tradeoff between accuracy rate and the number of fast matches required in the flow process of the server 130 and SGS 140. Thereafter, all the K patches are injected in parallel into all computational Cores 3 to generate K response vectors 22, which are fed into a signature generator system 23 to produce a database of Robust Signatures and Signatures 4.

In order to generate Robust Signatures, i.e., Signatures that are robust to additive noise L (where L is an integer equal to or greater than 1) by the Computational Cores 3 a frame T is injected into all the Cores 3. Then, Cores 3 generate two binary response vectors: S which is a Signature vector, and RS which is a Robust Signature vector.

For generation of signatures robust to additive noise, such as White-Gaussian-Noise, scratch, etc., but not robust to distortions, such as crop, shift and rotation, etc., a core C_i={n_i} (1≤i≤L) may consist of a single leaky integrate-to-threshold unit (LTU) node or more nodes. The node n_i equations are:

$V_i=\sum _j^{}{w}_{ij}{k}_j$ n _i =(Vi−Th_x)

where, is a Heaviside step function; w_ij is a coupling node unit (CNU) between node i and image component j (for example, grayscale value of a certain pixel j); k_j is an image component (for example, grayscale value of a certain pixel j); Thx is a constant Threshold value, where x is ‘S’ for Signature and ‘RS’ for Robust Signature; and Vi is a Coupling Node Value.

The Threshold values Thx are set differently for Signature generation and for Robust Signature generation. For example, for a certain distribution of Vi values (for the set of nodes), the thresholds for Signature (Th_S) and Robust Signature (Th_RS) are set apart, after optimization, according to at least one or more of the following criteria:

• • 1: For: V_i>Th_Rs 1−p(V>Th_S)−1−(1−ε)^l<<1

i.e., given that /nodes (cores) constitute a Robust Signature of a certain image I, the probability that not all of these I nodes will belong to the Signature of same, but noisy image, Ĩ is sufficiently low (according to a system's specified accuracy).

• • 2: p(V_i>Th_RS)≈l/Li.e., approximately l out of the total L nodes can be found to generate a Robust Signature according to the above definition.
  • 3: Both Robust Signature and Signature are generated for certain frame i.

It should be understood that the generation of a signature is unidirectional, and typically yields lossless compression, where the characteristics of the compressed data are maintained but the uncompressed data cannot be reconstructed. Therefore, a signature can be used for the purpose of comparison to another signature without the need of comparison to the original data. Detailed description of the Signature generation can be found U.S. Pat. Nos. 8,326,775 and 8,312,031, assigned to common assignee, which are hereby incorporated by reference for all the useful information they contain.

A Computational Core generation is a process of definition, selection, and tuning of the parameters of the cores for a certain realization in a specific system and application. The process is based on several design considerations, such as:

(a) The Cores should be designed so as to obtain maximal independence, i.e., the projection from a signal space should generate a maximal pair-wise distance between any two cores' projections into a high-dimensional space.

(b) The Cores should be optimally designed for the type of signals, i.e., the Cores should be maximally sensitive to the spatio-temporal structure of the injected signal, for example, and in particular, sensitive to local correlations in time and space. Thus, in some cases a core represents a dynamic system, such as in state space, phase space, edge of chaos, etc., which is uniquely used herein to exploit their maximal computational power.

(c) The Cores should be optimally designed with regard to invariance to a set of signal distortions, of interest in relevant applications. Detailed description of the Computational Core generation, the computational architecture, and the process for configuring such cores is discussed in more detail in the above-referenced U.S. Pat. No. 8,655,801.

The various embodiments disclosed herein can be implemented as hardware, firmware, software, or any combination thereof. Moreover, the software is preferably implemented as an application program tangibly embodied on a program storage unit or non-transitory computer readable medium consisting of parts, or of certain devices and/or a combination of devices. The application program may be uploaded to, and executed by, a machine comprising any suitable architecture. Preferably, the machine is implemented on a computer platform having hardware such as one or more central processing units (“CPUs”), a memory, and input/output interfaces. The computer platform may also include an operating system and microinstruction code. The various processes and functions described herein may be either part of the microinstruction code or part of the application program, or any combination thereof, which may be executed by a CPU, whether or not such a computer or processor is explicitly shown. In addition, various other peripheral units may be connected to the computer platform such as an additional data storage unit and a printing unit. Furthermore, a non-transitory computer readable medium is any computer readable medium except for a transitory propagating signal.

All examples and conditional language recited herein are intended for pedagogical purposes to aid the reader in understanding the principles of the invention and the concepts contributed by the inventor to furthering the art, and are to be construed as being without limitation to such specifically recited examples and conditions. Moreover, all statements herein reciting principles, aspects, and embodiments of the invention, as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Additionally, it is intended that such equivalents include both currently known equivalents as well as equivalents developed in the future, i.e., any elements developed that perform the same function, regardless of structure.

Claims

1. A method for detection of common patterns within unstructured data elements, comprising: searching a plurality of unstructured data elements extracted from big data sources to identify a plurality of patches;extracting the plurality of patches that were identified;generating, by a signature generator system, at least one signature for each one patch of the plurality of patches to generate a plurality of signatures for the plurality of patches, wherein the signature generator system includes a plurality of computational cores configured to receive the plurality of patches, each one computational core of the plurality of computational cores having properties that are at least partly statistically independent of other ones of the plurality of computational cores, wherein the properties of the one computational core are set independently of each other computational core of the plurality of computational cores;identifying common patterns among the plurality of signatures;clustering the plurality of signatures having the common patterns that were identified to generate a plurality of clusters; andcorrelating the plurality of clusters to identify associations between the respective common patterns that were identified.

2. The method of claim 1, further comprising: re-clustering clusters of the plurality of clusters having associations between their respective common patterns to create a common concept.

3. The method of claim 2, further comprising: storing, in a database, at least one of: the common concept, the plurality of signatures, the plurality of clusters, and the common patterns that were identified.

4. The method of claim 1, wherein the unstructured data elements are: specific interest unstructured data elements; or higher interest unstructured data elements than other unstructured data elements in the big data sources.

5. The method of claim 4, wherein each one of the specific interest unstructured data elements has an entropy level above a predefined threshold.

6. The method of claim 1, wherein the unstructured data elements are retrieved from a plurality of big data sources having similar classifications.

7. The method of claim 1, wherein each of the plurality of unstructured data elements is at least one of: a multimedia content, a book, a document, metadata, a collection of health records, audio, video, analog data, a file, unstructured text, or a web page.

8. The method of claim 1, wherein identifying the common patterns further comprises: matching the plurality of signatures to each other to generate a plurality of matched signatures;assigning a matching score to each pair of the plurality of matched signatures;comparing the matching score of each pair of the plurality of matched signatures to a preconfigured threshold; andupon determining that the matching scores of one pair of the plurality of matched signatures is above the predefined threshold, determining the one pair to have the common pattern.

9. The method of claim 1, wherein correlating the generated clusters further comprises: determining, for each one pair of the plurality of clusters, if the one pair contains at least a preconfigured number of matching signatures.

10. A non-transitory computer readable medium having stored thereon instructions for causing one or more processing units to execute the method according to claim 1.

11. A system for analyzing unstructured data, comprising: a network interface for allowing connectivity to a plurality of big data sources;a processing unit; anda memory connected to the processing unit, the memory containing instructions that, when executed by the processing unit, configure the system to: search a plurality of unstructured data elements extracted from the big data sources to identify a plurality of patches;extract the plurality of patches that were identified;generate, by a signature generator system, at least one signature for each one patch of the plurality of patches to generate a plurality of signatures for the plurality of patches, wherein the signature generator system includes a plurality of computational cores configured to receive the plurality of patches, each one computational core of the plurality of computational cores having properties that are at least partly statistically independent of other ones of the plurality of computational cores, wherein the properties of the one computational core are set independently of each other computational core of the plurality of computational cores;identify common patterns among the plurality of signatures;cluster the plurality of signatures having the common patterns that were identified to generate a plurality of clusters; andcorrelate the plurality of clusters to identify associations between the respective common patterns that were identified.

12. The system of claim 11, wherein the system is further configured to: re-cluster clusters of the plurality of clusters having associations between their respective common patterns to create a common concept.

13. The system of claim 12, wherein the system is further configured to: store, in a database, at least one of: the common concept, the plurality of signatures, the plurality of clusters, and the common patterns that were identified.

14. The system of claim 11, wherein the unstructured data elements are: specific interest unstructured data elements; or higher interest unstructured data elements than other unstructured data elements in the big data sources.

15. The system of claim 14, wherein each one of the specific interest unstructured data elements has an entropy level above a predefined threshold.

16. The system of claim 11, wherein the unstructured data elements are retrieved from a plurality of big data sources having similar classifications.

17. The system of claim 11, wherein each of the plurality of unstructured data elements is at least one of: a multimedia content, a book, a document, metadata, a collection of health records, audio, video, analog data, a file, unstructured text, or a web page.

18. The system of claim 11, wherein the system is further configured to: match the plurality of signatures to each other to generate a plurality of matched signatures;assign a matching score to each pair of the plurality of matched signatures;compare the matching score of each pair of the plurality of matched signatures to a preconfigured threshold; andupon determining that the matching scores of one pair of the plurality of matched signatures is above the predefined threshold, determine the one pair to have the common pattern.

19. The system of claim 11, wherein the system is further configured to: determine, for each one pair of the plurality of clusters, if the one pair contains at least a preconfigured number of matching signatures.