Automatic classification and interpretation of life science documents

Abstract

A computer-implemented tool for automatically classifying and interpreting documents, such as life science documents supporting clinical trials, the tool being configured to perform a combination of raw text analysis, document structure analysis, and image analysis to improve classification accuracy by achieving a more comprehensive machine-based understanding of document content. Compared to conventional automatic classification tools, the combination of analyses provides context for classification by exploiting relative spatial relationships between text and image elements, identifying features and formatting of elements, and extracting additional metadata from the document.

Description

Technical Field

[0001] The present invention relates to a computer-implemented method for classifying and interpreting life science documents and a computing device configured to be used as a computer-implemented automatic classification and interpretation tool; the present invention also relates to a non-transitory computer-readable storage medium. Background Art

[0002] Automatic classification is often used to effectively manage text, images, and unstructured information in documents. Automatic classification, also known as categorization, clustering, or text classification, typically enables documents to be divided and organized based on a set of predefined categories or classes for quick and easy retrieval of information. Summary of the invention

[0003] A computer-implemented tool for automatically classifying and interpreting documents, such as life science documents supporting clinical trials, is configured to perform a combination of text analysis, document structure analysis, and image analysis to improve classification accuracy by achieving a more comprehensive machine-based understanding of document content. Compared to traditional automatic classification tools, the combination of analyses provides context for classification by exploiting spatial relationships between text and image elements, identifying features and formats of elements, and extracting additional metadata from the document.

[0004] Text analysis identifies the original text and its general location in the document. In some implementations, text sequences can be selectively tracked. Document construction analysis provides additional understanding of the document to automatic classification and interpretation tools by identifying relevant context, such as spatial construction information (e.g., the location of images on the page; the location of text in headers, footers, and titles; the use of rows and columns in tables, etc.) and formatting (e.g., bold, italics, fonts, text size, etc.). Document construction analysis can also maintain connections between document elements by tracking text adjacent to document elements on either side. Image analysis targets graphic content, such as logos, diagrams, charts and tables, captions, etc., to extract additional context to further deepen the machine-based understanding of the document.

[0005] While text analysis can be used alone to achieve satisfactory results in some limited classification implementations, structural analysis and image analysis are not typically configured as separate operations for document classification. However, the combination of these three analyses can provide improved classification accuracy by operating in concert to help remove ambiguity between document categories, which in some cases may be numerous and subtly distinct. The combination of analyses distinguishes document categories by providing additional meaning using identified relevant context. This allows otherwise ambiguous document text to be effectively used for classification purposes.

[0006] In various illustrative examples, the automatic classification and interpretation tool is configured to apply machine learning to an artificial intelligence (AI) engine trained by historical classification documents. The AI ​​engine can be accessed by the tool, or incorporated into the tool in whole or in part, and used to implement combined text analysis, document structure analysis, and image analysis in whole or in part. A human operator can adjust the analysis algorithm through a user interface (UI) displayed on the automatic classification and interpretation tool. These adjustments can be used to improve the classification and interpretation of automatic documents, and can be used as machine learning input to improve the performance of the classification and interpretation tool as a whole.

[0007] The computing device supporting the automatic classification and interpretation tool may include an image capture device, such as a camera or scanner, or be configured to interoperate with a discrete image capture device for digitizing life science documents. The tool can be operated in the field to apply text analysis, document structure analysis, and image analysis to provide real-time classification feedback to human operators for captured documents. In some implementations, the real-time classification feedback may include a suggested classification for the captured document and associated metadata. The operator can review the suggested document classification and metadata and provide approval or corrections through the UI. Corrections can be used as machine learning input to improve the accuracy of the automatic classification and interpretation tool.

[0008] The computer-implemented automatic classification and interpretation tool provides improvements to the underlying operations of the computing device on which it is implemented by providing improved classification accuracy. More specifically, the combined use of raw text analysis, document structure analysis, and image analysis produces more accurate classifications by increasing the machine's knowledge depth to achieve efficient use of processing cycles, memory requirements, and network bandwidth, thereby reducing the need to perform manual reclassifications or discard incorrect classification results. The automatic classification and interpretation tool can also improve the efficiency of the human-machine interface on the computing device because the tool can more accurately classify life science documents, which improves the efficiency of subsequent computer-assisted document search and retrieval.

[0009] The present invention content is provided to introduce some concepts in a simplified form, and these concepts will be further described in the detailed description below. The present invention content is neither intended to identify the key features or essential features of the claimed subject matter, nor is it intended to be used to help determine the scope of the claimed subject matter. In addition, the claimed subject matter is not limited to the implementation of solving any or all shortcomings pointed out in any part of this disclosure. It will be appreciated that the above-mentioned subject matter can be implemented as a computer-controlled device, a computer processor, a computing system, or an article such as one or more computer-readable storage media. These and various other features will become apparent by reading the following detailed description and viewing the relevant drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 An exemplary computing environment is shown in which a computer-implemented automatic classification and interpretation tool is executed on a computing device;

[0011] Figure 2 An exemplary classification model is shown under which life science documents are classified as part of a trial master file (TMF);

[0012] Figure 3 Illustrative details of a computer-implemented automatic classification and interpretation tool executed on a computing device are shown;

[0013] Figure 4 shows illustrative data structures associated with analyses that are applied in combination during operation of a computer-implemented automatic classification and interpretation tool;

[0014] Figure 5 and Figure 6 An exemplary life science document that may be classified in TMF is shown;

[0015] Figure 7 Illustrative operational details of a text analysis process are shown;

[0016] Figure 8 Illustrative operational details of a construction analysis process are shown;

[0017] Fig. 9 Illustrative operational details of an image analysis process are shown;

[0018] Fig.10 is a flow chart showing an exemplary classification process;

[0019] Fig.11 shows an illustrative manual adjustment of a process operating on a computer-implemented automatic classification and interpretation tool, which can be used as machine learning input;

[0020] Fig.12 shows exemplary real-time classification feedback provided to an operator by an automated classification and interpretation tool;

[0021] Fig.13 , Fig.14 and Fig.15 An exemplary method is shown; and

[0022] Fig.16 is a block diagram of an illustrative computing device that may be used, at least in part, to implement the described automatic classification and interpretation of life science documents.

[0023] The same reference numerals indicate the same elements / components in the drawings. Unless otherwise indicated, the elements / components are not drawn to scale. DETAILED DESCRIPTION

[0024] Figure 1 An exemplary computing environment 100 is shown in which a human operator 105 employs a computing device 110 configured to support a computer-implemented automatic classification and interpretation tool 115. The tool may be implemented, for example, as a software application executed on a computing device. In alternative embodiments, the automatic classification and interpretation tool may be implemented using hardware, firmware, or a combination thereof, as desired for a particular implementation of the current automatic classification and interpretation of life science documents. The automatic classification and interpretation tool may interface with or directly include an artificial intelligence (AI) engine 120, which may be implemented, for example, using a neural or deep learning network, an expert system, or other suitable system configured for machine learning.

[0025] Computing device 110 includes electronic devices such as personal computers, servers, handheld devices, workstations, multimedia consoles, smart phones, tablet computers, laptop computers, etc. In the discussion that follows, the use of the term "computing device" is intended to encompass all electronic devices that perform some computing operations, whether they are implemented locally, remotely, or through a combination of local and remote operations.

[0026] The automatic classification and interpretation tool 115 can be configured to communicate with a remote computing device, resource, or service 130, optionally via a network link 125. Although this illustrative example employs processing at the local computing device 110 (as indicated by reference numeral 140), it is noted that other processing distributions and arrangements may also be used. For example, the automatic classification and interpretation tool may be instantiated as a remote or cloud-based application. For a given automatic classification and interpretation tool implementation, various combinations of local processing 140 and remote processing (as indicated by reference numeral 150) may be appropriately implemented.

[0027] Any of a variety of network types and network infrastructures may be used in various combinations or subcombinations, including local area networks (LANs), wide area networks (WANs), cellular networks, satellite networks, IP (Internet-Protocol) networks (such as Wi-Fi under IEEE 802.11 and Ethernet under IEEE 802.3), public switched telephone networks (PSTNs), and / or short-range networks (such as The communication network link 125 may be implemented by a network (e.g., a mobile operator, an enterprise, an Internet service provider (ISP), a telephone service provider, a data service provider, etc.). The communication network link 125 may utilize portions of the Internet (not shown) or include an interface supporting a connection to the Internet so that the computing device 110 can access data or content and / or provide a user experience supported by a remote service provider and / or other service providers (not shown).

[0028] Various image capture devices 155 may be utilized in a computing environment to capture images of life science documents that are deconstructed as described below. The image capture device may be a standalone device such as a camera or scanner, or may be incorporated into other computing devices 110. The image capture device may be deployed locally or remotely. For example, an image may be captured at a remote work site and transmitted to a local computing device 110 or to a cloud-based storage service and then accessed by the computing device over a network link. Alternatively, the computing device and the image capture device may be co-located.

[0029] Figure 2 An illustrative classification model 205 is shown under which life science documents 160 can be classified as part of a trial master file (TMF) 215. The life science industry is often required to comply with regulatory authorities in their oversight of the design, implementation, recording and reporting of clinical trials. Inspectors inspect the TMF to demonstrate that sponsors, trial monitoring personnel and researchers meet applicable standards and regulatory requirements. The TMF is a collection of life science documents that typically allows for the assessment of compliance with clinical trials and the quality produced. It should be noted that the term "life science documents" has general applicability to documents typically collected in a TMF.

[0030] The TMF 215 is organized with different categories or classes of life science documents 220...225. In some implementations, various subcategories 230 and content 235 may also be included in the TMF or used in classification systems other than those using the TMF. In some cases, the TMF may be organized to include categories or classes defined by the Drug Information Association (DIA).

[0031] Figure 3Illustrative details of a computer-implemented automatic classification and interpretation tool 115 executed on a computing device 110 are shown. The input to the tool includes life science documents 160, and the output from the tool includes a category label 330 and an event label 335 that are respectively applicable to each document in the document 160. The category label identifies the category 220...225 (or subcategory or appropriate content) that classifies the life sciences through the operation of the tool. The event label identifies objects in the document that the tool interprets as events, actions, and / or triggers that are meaningful. For example, the event label can identify document content related to security issues, regulatory factors, data management, and other information associated with the document content.

[0032] The computer-implemented automatic classification and interpretation tool 115 includes functionality configured for document deconstruction 305, text analysis 310, document structure analysis 315, image analysis 320, and classification and interpretation analysis 325. The classification and interpretation analysis may be supported by a classification model 205 and an interpretation model 340, which may be implemented as part of the tool 115 or externally as a database or resource on another computing device.

[0033] Document deconstruction 305 includes converting the captured life science documents 160 into a digitized form that uses a standardized data structure across all documents, for example, using XML (eXtensible Markup Languag) or other suitable representations to express the digitized form. In a typical implementation, the quality of the source material can be expected to vary greatly. Therefore, document deconstruction can apply various techniques to adapt to noise and undesirable artifacts during digitization to improve the quality of the input to the automatic classification and interpretation tool 115. In some cases, relevant descriptive information such as metadata can be collected for the input life science documents and stored. Such information can be used, for example, for TMF management and other purposes.

[0034] Figure 4 4 shows illustrative data structures respectively associated with text analysis, document structure analysis, and image analysis, which are applied in combination during operation of the computer-implemented automatic classification and interpretation tool 115. The life science document text structure 405 supports the text analysis 310 to achieve the extraction of text from the life science document 160 (generally without considering the sequence context (i.e., using a "bag of words" approach). Figure 1 ) to identify the original text. However, text sequences can be tracked and utilized in certain implementations of the tool.

[0035] The life sciences document library metadata structure 410 supports document construction analysis 315 to enable the utilization of document context, such as spatial construction information of document elements (e.g., location of images on the page; location of text in headers, footers, titles; use of rows and columns in tables, etc.) and formatting (e.g., bold, italics, fonts, text size, etc.). Document construction analysis can also maintain connections between document elements by tracking text adjacent to document elements on either side. The document image structure 415 supports image analysis 320 to enable graphical content such as logos, diagrams, charts and tables, captions, etc. to extract additional context to further deepen machine-based understanding of life sciences documents.

[0036] Illustrative examples of the application of text analysis, document structure analysis, and image analysis to life science documents are now presented. Figure 5 and Figure 6 Illustrative life science documents that may be classified in TMF are shown. Figure 5 An exemplary protocol summary 505 is shown, and Figure 6 An exemplary informed consent form 605 is shown. Details not relevant to the presentation of the example have been redacted from life science documents 505 and 605. The documents in this example are classified differently under DIA guidance, but each document has similar text content including the word "study." The following discussion illustrates how current computer-implemented automatic classification and interpretation tools and the application of a combination of text analysis, document structure analysis, and image analysis can achieve disambiguation and accurate classification of documents.

[0037] like Figure 7 As shown, text analysis 310 includes analysis of the original text without regard to formatting (as shown by reference numeral 705). As described above, in some implementations, the text sequence (710) can be selectively utilized to provide additional context. Here, for example, the word "research" identified by the text analysis is included in the informed consent document 605.

[0038] like Figure 8 As shown, document structure analysis 315 includes analyzing the relative spatial positions of document elements such as text and images (805). Element features (810) are also analyzed, including text font, size, and format (e.g., bold, italic, underline, subscript, superscript, etc.). Contextual connections (815) can also be analyzed by keeping the connection position of elements just before and after relative to the text information, using, for example, the concept of "nearest neighbor". Such content connections allow understanding the relevance of document elements (e.g., images) relative to their position within the document.

[0039] In this example, document structure analysis 315 identifies the word "study" in protocol summary document 505 as being located at the top of the first page of the document using a bold font. Document structure analysis 315 also identifies that the document structure includes the word located in a cell in a table with adjacent text "Study Title". Such features constitute metadata that provides additional context that enables computer-implemented automated classification and interpretation tools to gain a deeper understanding of protocol summary document 505 so that it is not inconsistent with informed consent document 605 ( Figure 6 ) produces blur.

[0040] like Fig. 9 As shown, image analysis 320 includes analyzing images in a document to generate additional metadata (905) that assists in document classification and interpretation. The location of the image in the document may also be identified (910). In some implementations, image-to-text conversion (915) may be applied to create a digitized representation of text elements that may be included in the image. In this example, image analysis identifies a handwritten "wet ink" signature 920 and date 925 to verify that the informed consent document has been signed. In some cases, the signature and date may be converted to a digital representation. Image analysis may also facilitate understanding of a graphic company logo 930 to enable further understanding of the entity that owns the document.

[0041] Fig.10 It is shown that it can be used for classification and interpretation analysis 325 ( Figure 3 ). In step 1005, relevant pages of life science documents of interest are collected. In step 1010, the document identification (ID) is checked. In step 1015, latent semantic analysis (LSA) is performed, which can weight document features (e.g., bold text, title location, logo, etc.) for relevance. In step 1020, the LSA results are collated into a nonparametric analysis of covariance (ANCOVA). In step 1025, maximum likelihood estimation is performed, and in step 1030, the documents are classified, for example, into the core DIA criteria.

[0042] Fig.11 A computer-implemented automatic classification and interpretation tool 115 ( Figure 1) is an illustrative manual adjustment 1105 of various text analyses, document structure analyses, and image analyses operated on the machine learning tool. In this example, the operator 105 can review 1115 the category labels 330 and event labels 335 output from the tool. The operator can adjust the document classification that the automatic classification and interpretation tool can analyze to make appropriate changes in the underlying automatic analysis. Alternatively, the operator itself can directly adjust the analysis to achieve the desired results. In some cases, the operator can perform multiple adjustment iterations to assist the machine learning process by specifying different classification results or by changing the processing parameters of each iteration.

[0043] Fig.12 An illustrative real-time classification feedback 1205 provided by the automatic classification and interpretation tool 115 to the operator 105 is shown. In this example, the tool may be instantiated solely on a portable computing device 110 that includes an image capture device, such as an integrated camera. Thus, the device 110 may be operated in an on-site location 1200 co-located with the life science documents 160, for example. Alternatively, the automatic classification and interpretation tool may be supported in part by a computing device 110 located in a home office or corporate enterprise, and / or using remote services 130 in a cloud computing configuration that uses various network links 1210 and 1215 to establish communication and interoperability between the various computing elements.

[0044] The automatic classification and interpretation tool 115 can be operated in a field location 1200 to quickly apply text analysis 310, document structure analysis 315, and image analysis and 320 to provide real-time classification feedback 1205 for captured documents to a human operator 105. In some implementations, the real-time classification feedback can include a suggested classification for the captured document and associated metadata. The operator can review the suggested document classification and metadata and provide approval or corrections through a UI displayed on the tool. The corrections can be used as machine learning input 1110 to improve the accuracy of the automatic classification and interpretation tool.

[0045] Fig.13 , Fig.14 and Fig.15 An illustrative method for automatic classification of life science documents is shown. Unless otherwise specified, the methods or steps shown in the flow chart and described in the accompanying text are not limited to a specific order or sequence. In addition, according to the requirements of such implementation, some of these methods or steps can occur or be performed simultaneously, and not all methods or steps must be performed in a given implementation, and some methods or steps can be selectively utilized.

[0046] Fig.13 can be supported by automatic classification and interpretation tools 115 ( Figure 1) is a flowchart of an illustrative method 1300 performed by a computing device of a computer program product. In step 1305, the device receives a digitized representation of a life science document, the digitized representation including a document element, the document element including one or more of text or an image. In step 1310, the device performs text analysis on the digitized representation of the life science document, the text analysis including identifying original words in the text. In step 1315, the device performs structural analysis on the digitized representation of the life science document, the structural analysis including identifying a document context, the document context describing the characteristics of the document element and the relative spatial position of the document element on the life science document page. In step 1320, the device performs image analysis on the digitized representation of the life science document, the image analysis including identifying an image and processing the identified image to extract additional features of the document element. In step 1325, the device collectively utilizes the results of the text analysis, structural analysis, and image analysis to classify the life science document into one or more predefined categories.

[0047] Fig.14 is supported by automatic classification and interpretation tools 115 ( Figure 1 ) is a flowchart of an illustrative method 1400 performed by a computing device of a computer program product. In step 1405, the device deconstructs one or more life science documents into a standardized data structure to generate document elements as input to a computer-implemented automatic classification and interpretation tool, the document elements including images and digitized text. In step 1410, the device performs a combination of text analysis, construction analysis, and image analysis on the document elements to create a context-based representation of the life science document, thereby identifying spatial relationships between the document elements. In step 1415, the device extracts metadata describing one or more of the document elements. In step 1420, the device uses the context-based representation and the extracted metadata to assist in classifying the life science document into predefined categories.

[0048] Fig.15 can be supported by automatic classification and interpretation tools 115 ( Figure 1 ) is a flowchart of an illustrative method 1500 performed by a computing device of a computer system. In step 1505, the device identifies original text in a digitized life science document. In step 1510, the device identifies the structure of the digitized life science document to identify the relative spatial positions of text and image elements in the digitized life science document. In step 1515, the device identifies an image to extract text in digitized form. In step 1520, the device identifies features of the original and extracted text. In step 1525, the device combines the results of each of the identified steps to generate metadata. In step 1530, the device classifies the life science document using the generated metadata.

[0049] Fig.16 An exemplary architecture 1600 is shown for a device (eg, a server) capable of executing the various components described herein for automated classification and interpretation of life science documents. Fig.16 The architecture 1600 shown in FIG. 1 includes one or more processors 1602 (e.g., central processing units, dedicated artificial intelligence chips, graphics processing units, etc.), system memory 1604 including RAM (random access memory) 1606 and ROM (read-only memory) 1608, and a system bus 1610 that operably and functionally couples the components in the architecture 1600. A basic input / output system containing basic routines that help to pass information between elements in the architecture 1600 (e.g., during startup) is typically stored in ROM 1608. The architecture 1600 also includes a mass storage device 1612 for storing software code or other computer-executed code used to implement applications, file systems, and operating systems. The mass storage device 1612 is connected to the processor 1602 through a mass storage controller (not shown) connected to the bus 1610. The mass storage device 1612 and its associated computer-readable storage media provide non-volatile storage for the architecture 1600. Although the descriptions of computer-readable storage media contained herein refer to mass storage devices, such as hard disks, solid-state drives, or optical drives, it will be understood that computer-readable storage media may be any available storage media accessible by architecture 1600 .

[0050] By way of example and not limitation, computer-readable storage media may include volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules, or other data. For example, computer-readable media include, but are not limited to, RAM, ROM, EPROM (erasable programmable read-only memory), EEPROM (electrically erasable programmable read-only memory), flash memory or other solid-state storage technology, CD-ROM, DVD, HD-DVD (high-definition DVD), Blu-ray or other optical storage, cassettes, magnetic tape, disk storage devices or other magnetic storage devices, or any other medium that can be used to store the desired information and that can be accessed by the architecture 1600.

[0051] According to various embodiments, the architecture 1600 can operate in a network environment using logical connections to remote computers through the network. The architecture 1600 can be connected to the network through a network interface unit 1616 connected to the bus 1610. It is understood that the network interface unit 1616 can also be used to connect to other types of networks and remote computer systems. The architecture 1600 can also include an input / output controller 1618 for receiving and processing input from a number of other devices, including a keyboard, a mouse, a touch pad, a touch screen, a control device such as a button and switch or an electronic pen ( Fig.16 Similarly, input / output controller 1618 may provide output to a display screen, user interface, printer, or other type of output device (also not shown). Fig.16 ).

[0052] It is understood that when the software components described herein are loaded into the processor 1602 and executed, the processor 1602 and the entire architecture 1600 can be transformed from a general-purpose computing system to a customized special-purpose computing system to facilitate the functions proposed herein. The processor 1602 can be composed of any number of transistors or other discrete circuit elements, which can present any number of states individually or collectively. More specifically, the processor 1602 can operate as a finite state machine in response to executable instructions contained in the software modules disclosed herein. These computer executable instructions can transform the processor 1602 by specifying how the processor 1602 switches between states, thereby transforming the transistors or other discrete hardware elements that constitute the processor 1602.

[0053] Encoding the software module proposed herein can also transform the physical structure of the computer-readable storage medium proposed herein. In different implementations of this specification, the specific transformation of the physical structure can depend on various factors. Examples of these factors may include, but are not limited to: the technology used to implement the computer-readable storage medium, whether the computer-readable storage medium is characterized as primary storage or secondary storage, etc. For example, if the computer-readable storage medium is implemented as a semiconductor-based memory, the software disclosed herein can be encoded on the computer-readable storage medium by transforming the physical state of the semiconductor memory. For example, the software can transform the state of the transistors, capacitors, or other discrete circuit elements that constitute the semiconductor memory. The software can also transform the physical state of these components to store data thereon.

[0054] As another example, magnetic or optical technology can be used to implement the computer-readable storage medium disclosed herein. In such an implementation, when the software is encoded in a magnetic or optical medium, the software proposed herein can transform the state of the magnetic or optical medium. These transformations can include changing the magnetic characteristics of a specific location in a given magnetic medium. These transformations can also include changing the physical characteristics or properties of a specific location in a given optical medium to change the optical properties of those locations. Without departing from the scope and spirit of this specification, other transformations of physical media are possible, and the aforementioned examples are provided only to facilitate this discussion.

[0055] In view of the above, it can be appreciated that many types of physical transformations occur in architecture 1600 in order to store and execute the software components presented herein. It can also be appreciated that architecture 1600 can include other types of computing devices, including wearable devices, handheld computers, embedded computer systems, smart phones, PDAs, and other types of computing devices known to those skilled in the art. It is also contemplated that architecture 1600 may not include Fig.16 All components shown can be included in Fig.16 Other components not explicitly shown in the drawings or may be used in conjunction with Fig.16 The architecture shown is a completely different architecture.

[0056] The subject matter described above is provided by way of example only and should not be construed as limiting. Various modifications and changes may be made to the subject matter described herein without following the example implementations and applications shown and described, and without departing from the true spirit and scope of the invention as set forth in the following claims.

Claims

1. A computer-implemented method for classifying and interpreting life science documents, the method comprising: receiving a digitized representation of the life science document, the digitized representation comprising document elements including one or more of text or images; performing text analysis on the digitized representation of the life science document, the text analysis comprising identifying original words in the text; performing a structural analysis on the digitized representation of the life sciences document, the structural analysis comprising identifying a document context, the document context describing characteristics of document elements and relative spatial locations of document elements on a page of the life sciences document; performing image analysis on the digitized representation of the life science document, the image analysis comprising recognizing an image and processing the recognized image to extract additional features of document elements; collectively utilizing the results of the text analysis, the construct analysis, and the image analysis to classify the life science document into one or more predefined categories; as well as The life science document is tagged with a classification tag and an event tag, wherein the event tag is configured to operate as a trigger or an alarm.

2. The computer-implemented method of claim 1 , wherein: The relative spatial position includes one of a header, a footer, a caption, a footnote, or a title.

3. The computer-implemented method of claim 1 , wherein: Identifying the context also includes identifying the format of the life science document.

4. The computer-implemented method of claim 1 , wherein: The image analysis may also include identifying logos, graphics, illustrations, illustrated text or captions.

5. The computer-implemented method of claim 4, wherein: The image analysis also includes interpreting one or more of the identified logo, the graphic, the illustration, the illustrated text, or the caption.

6. The computer-implemented method of claim 1 , wherein: The characteristic of the document element includes one of a font, size, or format of the text.

7. The computer-implemented method of claim 1 , wherein: The structural analysis also includes tracking the text of adjacent document elements.

8. The computer-implemented method of claim 1 , wherein: The image analysis also includes image to text conversion to extract text in digitized form from the image.

9. The computer-implemented method of claim 1, further comprising categorizing content in the life science document into one or more predefined categories.

10. The computer-implemented method of claim 1, wherein: The text analysis includes tracking a sequence of text in the life sciences document.

11. The computer-implemented method of claim 1 , wherein: One or more of the text analysis, the construction analysis, or the image analysis generates metadata associated with the life science document, wherein the metadata is used at least in part to perform the classification.

12. The computer-implemented method of claim 1, wherein: The one or more predefined categories include categories defined by a drug information association.

13. A computing device configured to operate as a computer-implemented automatic classification and interpretation tool, the computing device comprising: one or more processors; and one or more non-transitory computer-readable storage media storing instructions that, when executed by the one or more processors, cause the computing device to: deconstructing one or more life science documents into a standardized data structure to generate document elements including images and digitized text as input to a computer-implemented automated classification and interpretation tool, performing a combination of text analysis, structural analysis, and image analysis on the document elements to create a context-based representation of the life science document to identify spatial relationships between the document elements, extracting metadata describing one or more of the document elements, utilizing the context-based representation and the extracted metadata to assist in classifying the life science documents into predefined categories; The life science document is tagged with a classification tag and an event tag, wherein the event tag is configured to operate as a trigger or an alarm.

14. The computing device of claim 13, wherein: The instructions when executed further cause the computing device to interpret the life science document as an event, action, or trigger.

15. The computing device of claim 13, wherein: The executed instructions further cause the computing device to classify the life science document using a machine learning process, wherein the machine learning process is adjustable based on input from a human operator.

16. One or more non-transitory computer-readable storage media storing executable instructions that, when executed by one or more processors in a computing device, implement a computer-implemented automatic classification tool configured to perform a method comprising the steps of: Identify original text in digitized life science documents; identifying the structure of the digitized life science document to identify the relative spatial positions of text and image elements in the digitized life science document; Recognize images to extract text in digitized form; Identify features of original and extracted text; Combining the results of each of the steps of identifying to generate metadata; classifying the life science document using the generated metadata; as well as The life science document is tagged with a classification tag and an event tag, wherein the event tag is configured to operate as a trigger or an alarm.

17. The one or more non-transitory computer-readable storage media of claim 16, wherein: The classification utilizes one or more of weighting the results, application of latent semantic analysis, or non-parametric ANCOVA (analysis of covariance).

Images