System and procedure for identifying and / or extracting information relevant for submitting a bid from a document relating to a tender or request for quotation

The system automates the extraction of relevant information from tender documents using networked modules for typing, text recognition, and keyword search, addressing inefficiencies in manual document review and enhancing document readability and storage efficiency.

DE102019122223B4Active Publication Date: 2024-11-28CORTEX MEDIA GMBH
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
DE102019122223
Authority / Receiving Office
DE · DE
Patent Type
Patents
Current Assignee / Owner
Filing Date
2019-08-19
Publication Date
2024-11-28
Estimated Expiration
2039-08-19

AI Technical Summary

Technical Problem

Existing methods for identifying relevant information in extensive tender or inquiry documents are inefficient, often requiring manual search through non-digital or non-text-based documents, leading to errors and time consumption, especially in large projects.

Method used

A system and method utilizing a network-connected evaluation computer with modules for typing, text recognition, search, and output to automatically identify and highlight keywords in machine-readable documents, including optical character recognition and regular expressions, to generate a highlighted or reduced-size output document.

Benefits of technology

Facilitates efficient and accurate extraction of relevant information from tender documents, reducing time and effort, and minimizing errors, while enabling storage optimization and improved document readability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

System (100) for identifying and / or extracting information relevant for submitting a bid from a document relating to a tender or request for proposals, with a network (103) that is designed to connect an evaluation computer (104) with at least one provider computer (101) for the exchange of digital documents (200, 201) at least temporarily, with a transmit and / or receive module (102) assigned to the provider computer (101), which is configured to send digital documents (200) via the network (103) to the evaluation computer (104) or to retrieve such documents via the network (103), and with a transmit and / or receive module (105) assigned to the evaluation computer (104), which is configured to provide digital documents (200, 202) via the network (103) for retrieval by the provider computer (101) or to receive such documents via the network (103), where the evaluation computer (104) has: - a typing module (106) trained to type a digital document (200) provided to it as being in the format of the extended markup language (XML), a text-based PDF document, a vector-based machine-readable document, or a raster-based machine-unreadable document, wherein the digital document (200) is first typed based on its file extension, and wherein, to avoid incorrect file extensions of the digital document (200), a preliminary analysis is performed in which, in the case of PDF documents, an attempt is made to extract text, and only if a threshold for the number of letters or characters is exceeded, page by page or in the entire document, is it concluded that it is a text-based machine-readable PDF document, wherein, if the threshold is not reached, it is concluded that it is an image-based PDF document as a machine-unreadable document, - a text recognition module (107) that is trained to detect letters present in a digital document (200) previously typed as machine-unreadable and thus to generate a machine-readable digital document (200) from the machine-unreadable document, - a search module (108) that is configured to search the machine-readable digital document (200) for at least one keyword and / or for at least one keyword variation generated on the basis of a regular expression, wherein the at least one keyword and / or the at least one regular expression are stored in a database (109) of the evaluation computer (104), as well as - an output module (110) that is configured to produce a digital output document (202) in which the at least one keyword found and / or the at least one keyword variation found is highlighted and / or marked.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] The invention relates to a system and a method for identifying and / or extracting information relevant for submitting a bid from a document relating to a tender or a request for proposals, particularly a lengthy one. The invention also relates to a computer program implementing this method.

[0002] In many industries, it is common practice to send requests for proposals to a large number of providers in order to select the most advantageous one from a range of offers. Furthermore, particularly when public bodies award contracts, projects are announced through a call for tenders, for which providers can then apply or submit a bid. Such request for proposals or tender documents can be very extensive, depending on the project. A lengthy document, for example, may contain more than ten pages. Therefore, especially in large-scale projects, providers of individual trades usually only have a fraction of the pages contained in the request for proposals or requests for proposals that are relevant. It is very time-consuming to search through the entire document for the relevant sections or tender sections.A computer-assisted search for the relevant pages of a tender or request for proposals is still highly error-prone, as the documents are either not available digitally or, if they are available digitally, are not text-based and therefore cannot be searched. At the same time, the non-searchable documents are often of insufficient quality due to their grid format. Furthermore, the diverse nature of requests for proposals or tenders makes it difficult for providers to quickly locate the information they are interested in.

[0003] US 2008 / 0 246 998 A1 describes a method for coloring monochromatically scanned documents. The coloring is performed based on coloring rules that, for example, specify the coloring based on predefined fonts. In one of the embodiments described therein, it is possible to automatically color those text passages that were previously stored as keywords in a database.

[0004] US Pat. No. 5,825,943 A describes a method and system for digitizing and subsequently recognizing text in a digitized document. This involves initially generating two data sets for a physical document: an image data set and a text data set, with the text data set resulting from OCR recognition. Based on this data, a table is then created that provides a reference between the images and the text for later browsing. A user can then search for specific terms, even identifying entire phrases that appear across page breaks in the document.

[0005] US 2006 / 0 062 453 A1 describes a method for automatically marking or highlighting text passages in a document. This involves first manually highlighting a text passage on a physical document with a highlighter, and then scanning the physical document. The automatic processing of this document then searches for further text passages that correspond to the previously manually highlighted terms.

[0006] It is therefore the object of the present invention to provide a system and a method for identifying and / or extracting information relevant for submitting a bid from a lengthy document relating to a tender or a request for proposal, which takes into account at least one of the aforementioned disadvantages. It is also the object of the present invention to provide a corresponding computer program.

[0007] This object is achieved with a system having the features of claim 1, with a method having the features of claim 7 and with a computer program having the features of claim 10. Advantageous embodiments with expedient further developments of the invention are specified in the dependent claims.

[0008] The system according to the invention is particularly equipped with a network designed to at least temporarily connect an evaluation computer to at least one provider computer for the exchange of digital documents. The system comprises a transmitting and / or receiving module assigned to the provider computer, which is designed to send digital documents to the evaluation computer via the network or to retrieve such documents from the evaluation computer via the network. In addition, a transmitting and / or receiving module assigned to the evaluation computer is provided, which is designed to provide digital documents via the network for retrieval by the provider computer or to receive such documents via the network. The evaluation computer has, in particular: - a typing module designed to type a digital document submitted to it as a machine-readable document or a machine-unreadable document, - a text recognition module designed to detect letters present in a digital document previously typed as machine-unreadable and thus to generate a machine-readable digital document from the machine-unreadable document, - a search module designed to search the machine-readable digital document for at least one keyword and / or for at least one keyword variation generated on the basis of a regular expression, wherein the at least one keyword and / or the at least one regular expression are stored in a database of the evaluation computer, and - an output module configured to generate a digital output document in which the at least one retrieved keyword and / or the at least one retrieved keyword transformation is highlighted and / or marked.

[0009] The system has the advantage that a request for proposals or a tender can be automatically searched for the information relevant to submitting a bid, whereby the output document preferably corresponds to the digital document originally submitted to the evaluation computer, but is supplemented by comments or highlights at those points where the keyword found or where the keyword modification resulting from the use of the regular expression is present.

[0010] A bidder computer is the computer or system component that can be operated by a user who wishes to participate in a tender or respond to an inquiry by submitting a bid. However, the bidder computer can also act as a server, which is then not necessarily operated by a human operator but can also operate automatically. An evaluation computer is the computer or system component that processes, searches, and prepares a document relating to the tender or inquiry in order to quickly provide the user with the information relevant to submitting a bid.

[0011] The use of the network provides a wide variety of communication options between the provider computer and the evaluation computer. For example, it is possible for a server-client communication connection to be active, at least temporarily, between the evaluation computer and the provider computer. In this case, for example, the evaluation computer would be configured as a server, in particular as an application server, equipped with a transmitting and receiving module to receive a digital document from a constituent of the system and then send the processed output document. The provider computer is configured, for example, as a client, which can also be equipped with only a receiving module to receive the output document.Furthermore, it is possible for both the evaluation computer and the provider computer to be in a server-to-server communication connection, for example if the provider computer itself has a corresponding document management system for managing and storing digital documents, to which the evaluation computer has access. For example, the exchange of digital documents and / or output documents can take place via the network using email, so that a mail server is available either on the evaluation computer itself or in the network to provide the evaluation computer with corresponding incoming inquiries or tenders. It is also possible for the evaluation computer to provide a portal for uploading digital documents in order to subsequently process the digital document and then make the output document available for download. Alternatively or in addition, many other interfaces (e.g.TCP / IP, UDP, IMAP, HTTP, HTTPs, FTP, FTPs, SCP, SSH, XML, etc.) must be present to allow the digital document to be transported over the network. The network also enables connection to a document management system, especially from third-party computers.

[0012] However, it is also possible for the document, especially a large one, to be in the form of a physical or digital fax document or a physical document sent by post, for example in paper form. For such cases, the system preferably has a corresponding scanning device to generate a digital document from the physical document. In this case, the scanning device can, on the one hand, automatically perform the corresponding letter recognition so that a machine-readable document can be made available to the evaluation computer. If, on the other hand, the scanning device can only generate image-based digital documents, the system according to the invention first provides the corresponding text recognition in its text recognition module of the evaluation computer in order to subsequently generate a machine-readable digital document.

[0013] It has proven advantageous if the output document is an extraction document that is reduced in size compared to the machine-readable digital document and contains at least those pages containing the at least one retrieved keyword and / or the at least one retrieved keyword variation. This eliminates the need for a provider to review the entire original document to access the relevant text passage, as only those pages containing the relevant information are provided. Furthermore, such a reduction and provision of the extraction document reduces storage requirements.

[0014] According to the invention, a document in portable document format, which is text-based and thus searchable (so-called "true PDF"), can be used as a machine-readable electronic document. In general, "Portable Document Format" (PDF) is a file format developed by Adobe Systems Inc. that allows PDF-formatted documents to be accessed for online viewing, for example, using a computer screen, in the same way that printouts of the documents could be viewed offline. For example, PDF documents can be viewed page by page or scrolled online, depending on the user's specific needs.The content of the PDF document can alternatively consist only of images or be in image form (so-called “Image-only Portable Document Format”), so that it is not text- or vector-based but raster-based and therefore represents a machine-unreadable document.

[0015] According to the invention, a document can be presented in extensible markup language (XML) format. This document format has proven to be an excellent machine-readable electronic format. In the construction industry in particular, a possible proposal or standard for data exchange based on such XML documents exists. The so-called GAEB XML format (GAEB stands for "Joint Committee for Electronics in Construction") has proven useful for efficient tendering, efficient contract awarding, and efficient invoicing of construction services. This is an exchange format suitable for the construction industry for machine-reading the relevant information for submitting bids.It has therefore proven to be useful for evaluation because with a digital document in XML format, no individual pages and therefore no page breaks need to be handled and, at the same time, sections are clearly defined via the XML structure.

[0016] The typing module first types the digital documents received at the evaluation computer based on their file extension. However, it can happen that, for example, a document was typed as an XML document and thus as machine-readable because someone specified an "incorrect" file extension for the digital document. To counteract this problem, the invention performs a detailed preliminary analysis of the content of the digital document. For PDF documents, an attempt is first made to extract the text from the PDF. Metadata can also be helpful in this process to type the PDF accordingly. If a predefined threshold for the number of letters or characters in the entire document is exceeded, it is concluded that it is a text-based (true) PDF document. If the threshold is not reached or exceeded, it is considered an image-based PDF document.During the review process, a dictionary comparison can help to check whether the text of the PDF document can be considered meaningful in order to more clearly classify it as a machine-readable, digital document.

[0017] The above-described check in the typing module can also be performed page by page. This check checks whether a threshold for the number of characters or the number of meaningful words has been exceeded, in order to obtain an indication of the presence or absence of a machine-readable document.

[0018] The typing module is preferably also configured to perform text recognition on the images contained in the PDF document. This can determine the number of letters and / or characters present in the examined images. Calculating the difference between the number of characters in the document before scanning the images and the number of characters present in the document, together with the number of characters in the images, also provides an indicator of whether the document is machine-readable or machine-unreadable. A corresponding threshold can also be specified for this difference.

[0019] The images contained in the PDF document can, on the one hand, contain text, i.e. be text blocks in image form, or on the other hand, be "real images" in the form of graphics, logos, photos, figures, illustrations or sketches. Text recognition, for example, can always be carried out, although after recognition it could also be determined that the PDF document only contains images without any text blocks. It has proven efficient and resource-saving to only subject those images to recognition that contain text blocks in image form in order to extract the text they contain. This does not take into account, for example, images that are always in the same place with the same pixels on each page. This means that based on this information it can be concluded that the image is probably a logo or a constant header of a PDF document.A threshold for the occurrence of such an image can also be set for the evaluation.

[0020] The text recognition module is preferably designed to add a machine-readable layer to machine-unreadable (PDF) documents in order to generate the machine-readable digital document to be subsequently searched. However, to further improve character recognition, it has proven advantageous if the machine-readable document is in the form of an image-based portable document format and if the text recognition module includes an image optimization device for graphically pre-processing the document in the image-based portable document format, which in turn is connected upstream of an optical character recognition device of the text recognition module.

[0021] The image optimization device preferably has an alignment unit for aligning one or more of the pages of the document by rotating them so that subsequent text recognition can be carried out more reliably. Rotation is understood to mean any rotation of the document or of individual pages of the document between 0 degrees and 360 degrees, whereby the individual pages of the digital document often only need to be rotated by a maximum of 10 degrees in order to correct the alignment of a document that has been fed in at an angle and is therefore scanned "slanted". In addition, it has proven advantageous if the image optimization device has a contrast sharpening unit designed to strengthen and / or sharpen the existing contrasts, so that character recognition in the optical character recognition device is also carried out more reliably.

[0022] The optical character recognition (OCR) system is primarily dictionary-based. The optical character recognition system is designed to first recognize the language of the digital document and then select the dictionary corresponding to the recognized language for text recognition. During character recognition, a comparison is then made with the words contained in the selected dictionary, resulting in improved and more efficient character recognition.

[0023] In an advantageous further development, the optical character recognition device is assigned an artificial neural network that is designed and trained to recognize the correct word from the context despite individual unrecognizable letters. This artificial neural network preferably has recurrent wiring to provide even better character / word recognition. In this context, it is advantageous if the artificial neural network also has an LTSM (long-term short memory) structure, so that a larger amount of data can be or could be accessed during training of the artificial neural network, which contributes to even better text / word recognition.

[0024] During optical character recognition, it may happen that the page information of the digital document is lost, making it impossible to reliably identify the areas or pages containing information relevant for submitting a bid. For this reason, it is advantageous to have a separating and / or combining device interposed between the image optimization device and the optical character recognition device. This device is configured to provide the pages of the digital document to the optical character recognition device individually or to combine individual pages back into the entire digital document.Thus, the optical character recognition device can subject the pages of the digital document to the recognition procedure page by page, after which the assembled complete document with the corresponding page information is then available for further analysis.

[0025] Another essential core component of the system is the search module, which is at least temporarily connected to a database. Preferably, several keywords are stored in the database. Preferably, a plurality of regular expressions are also stored in the database so that keyword modifications or variants of keywords can be found during a search. The use of regular expressions offers a powerful tool for the reliable recognition of information relevant to submitting an offer, which is only discussed in more detail in the figure description as an example. The search module is preferably designed to take a predefined or predefinable Hamming distance into account when evaluating the character strings during the search for digital documents that were originally typed as machine-unreadable but were then made machine-readable.The Hamming distance can be one, two, or more than two, with the correction of individual characters in the character strings being performed according to the principle of probability, if necessary also by comparing them with words from a digital dictionary. It is also possible to apply the Hamming distance to digital documents typed as machine-readable. Since errors in such documents are more likely to be due to typos in the original document, rather than to possible misrecognition during text recognition, it is preferable that the search taking into account the Hamming distance (correction using) is performed exclusively on digital documents originally typed as machine-unreadable.

[0026] For the purposes of the present invention, the calculation of a Hamming distance can be omitted when using the artificial neural network explained above. Conversely, it is possible to omit the artificial neural network but still calculate the Hamming distance. Furthermore, it is also possible to use or employ both the artificial neural network and the calculation of the Hamming distance.

[0027] It is useful if the search module or the text recognition module is designed to recognize text paragraphs or sections or chapters with identified keywords and / or identified keyword variations in the digital document, and if the output module is designed to output the pages of the digital document with the at least one complete text paragraph or with the at least one complete section or chapter in the output document in which the at least one found keyword and / or the at least one found keyword variation is contained. The recognition is carried out block by block, for example by searching for numbers and text behind the numbers. Such a structure of number and text is typically an indication of the beginning of a new chapter or a new section, so that the digital document can be divided into individual blocks in a preliminary examination.If a keyword or keyword variation is found in a block, the block is retained in its entirety during output, even if there is a page break. In other words, the system ensures that the entire section and chapter blocks are retained even across the page break, so that a bidder receives all relevant information related to the found keyword or keyword variation in the subsequent output or extraction document, enabling them to subsequently create and submit their bid.

[0028] The system, in particular the evaluation computer, can also have a document generation module connected in parallel to the text recognition module, which, for example, creates a machine-readable document in a second format from a machine-readable digital document in a first format. Thus, for example, a digital document recognized as an XML file could also be converted into a (true) PDF document by the document generation module before it is forwarded to the search module. Converting the XML file to PDF format is advantageous because, due to the structure of the XML document, which is formed by markup of text sections, it is very difficult for a processor to identify the information relevant to them in the tender or inquiry. By converting it to a PDF document, the readability of the content of the original XML document is improved for the processor, or even created in the first place.

[0029] The procedure for identifying and / or extracting information relevant for submitting a tender from a lengthy document relating to a call for tenders or a request for proposals includes, in particular, the following steps: - Providing a digital document to a computer or to an evaluation computer; - Typing the digital document using a typing module to determine whether the digital document is a machine-readable document or a machine-unreadable document; - Creating a machine-readable digital document using a text recognition module if the document was previously typed as machine-unreadable; - Searching the machine-readable document by means of a search module for at least one keyword relevant to the offer and / or for at least one keyword variation generated on the basis of a regular expression, wherein the at least one keyword and / or the at least one regular expression is stored in a database of the evaluation computer; - generating a digital output document by means of an output module in which the at least one found keyword and / or the at least one found keyword variation is highlighted and / or marked; and - Providing or outputting the output document by the computer or the evaluation computer.

[0030] The advantages and advantageous embodiments mentioned in connection with the system apply equally to the method according to the invention, which is characterized by a simplified retrieval of the information relevant for submitting an offer.

[0031] Since the individual process steps are computer-implemented, it can be implemented as a native solution, i.e., without a network connection; it can also be designed and implemented as an email solution, a server-server solution, or as a portal solution with a client-server structure within a network. The various communication options have already been discussed above.

[0032] The machine-unreadable document is often in an image-based portable document format, which is why it is advantageous if the text recognition module comprises an image optimization device to pre-process the document graphically in the image-based portable document format. The image optimization device has an alignment unit that aligns one or more of the pages of the document by means of rotation. The image optimization device also has a contrast sharpening unit that enhances and / or sharpens the existing contrasts. The rotation can be only a few degrees; for example, a maximum of 10 degrees.The text recognition module further comprises a device for optical character recognition, which recognizes the letters of the previously graphically prepared document, in particular dictionary-based and in particular with the aid of an artificial neural network. A machine-readable layer is added to the machine-unreadable digital document in order to generate a machine-readable digital document from the machine-unreadable digital document. The machine-readable layer can typically be marked by a user, but it remains invisible to them. In other words, the machine-readable layer is text superimposed over the image with an invisibly displayed font color.

[0033] These procedural steps can reduce subsequent complications during text recognition, which may arise due to, for example, insufficient resolution of the machine-unreadable document, insufficient bit depth, skewed scanning, insufficient contrast, or a noisy background. Such complications can lead to unwanted misrecognitions.

[0034] The method also has the advantage that the pages of the digital document are separated by means of a separating and / or combining device prior to optical character recognition. Preferably, the separated pages are recombined into a complete, machine-readable digital document after character recognition by means of the separating and / or combining device. This has the advantage that the information contained in the pages is not lost during optical character recognition, as this information can be retained by the separating and / or combining device.

[0035] The computer program according to the invention has the same advantages and advantageous configurations as the method and system according to the invention. It has proven particularly advantageous when loaded into a computer's digital memory and implementing the corresponding method.

[0036] The features and feature combinations mentioned above in the description, as well as the features and feature combinations mentioned below in the description of the figures and / or shown alone in the figures, can be used not only in the respective combinations specified, but also in other combinations or on their own, without departing from the scope of the invention. Thus, embodiments are also to be considered encompassed and disclosed by the invention that are not explicitly shown or explained in the figures, but which emerge and can be produced from the explained embodiments through separate feature combinations.

[0037] Further advantages, features, and details of the invention will become apparent from the claims, the following description of preferred embodiments, and the drawings. These show: Fig. 1 a schematic system for identifying and / or extracting information relevant for the submission of a tender from a document relating to a call for tenders or a request for proposals; Fig. 2 a second schematically illustrated system, wherein a tendering computer is shown and wherein the evaluation computer additionally has a document generation module; Fig. 3 a schematic detailed view of the text recognition module, Fig. 4 an example database of the evaluation calculator with example keywords and example regular expressions, and Fig. 5 a schematic representation of the procedure.

[0038] In the Fig. 1 and Fig. 2 shows a system 100 for identifying and / or extracting information relevant to submitting a bid from a comprehensive document relating to a tender or an inquiry. An evaluation computer 104 is connected, at least temporarily, to at least one provider computer 101 for exchanging digital documents 200, 201, 202 via a network 103. A plurality of provider computers 101 may also be present (another is shown in dashed lines here) in order to connect to the evaluation computer 104, which is configured in particular as a dedicated server, via the network 103. Each provider computer 101 has a transmitting and / or receiving module 102 configured to transmit digital documents 200 to the evaluation computer 104 via the network 103 and / or to retrieve such documents via the network 103.The evaluation computer 104 also has a transmitting and / or receiving module 105, which is designed to provide the digital documents 200, 202 via the network 103 for retrieval by one of the provider computers 101 and / or to receive such documents via the network 103.

[0039] The evaluation computer 104 itself comprises, in addition to its transmitting and / or receiving module 105, a typing module 106, a text recognition module 107, a search module 108, a database 109 and an output module 110. The functioning of the individual modules and the method according to the invention are explained below.

[0040] First, the evaluation computer 104 receives a digital document 200 via its transmitting and / or receiving module 105. This digital document 200 can be retrieved, for example, via a mail server assigned to the system 100, whereby the evaluation computer 104 itself can also have a corresponding mail module. It is also possible for the user on the provider computer 101 to log in to the evaluation computer 104 via a portal in order to upload the digital document 200 via the network 103.

[0041] In the system 100 of Fig. 2 is in contrast to the system 100 according to Fig. 1, a tendering computer 111 is additionally present, which also includes a transmitting and / or receiving module 112, via which the tendering computer 111 can communicate with the network 103. The tendering computer 111 is understood to be the computer or constituent of the system 100 that is operable by the user who creates the tender or the request for quotation and makes it available to potential suppliers. The tender or the request for quotation is created on the tendering computer 111 and, if necessary, sent as an image-based or text-based digital document 200 via the network 103 to the evaluation server 104. Alternatively or additionally, the digital document 200 can be sent via email to the supplier computer 101, so that the supplier computer then uploads the digital document to the evaluation computer 104, where it is checked for its relevance for submitting a tender.

[0042] If the extensive document is initially available in paper form, it is digitized using a suitable scanning device, resulting in a comprehensive digital document 200. The digital document 200 is then transferred to the typing module 106, which checks the digital document 200 to determine whether it is a machine-readable document, for example, a GAEB XML document or a true PDF document, or whether it is a machine-unreadable document, for example, an image-based PDF document.

[0043] A document classified as machine-unreadable is transferred to the text recognition module 107, which creates a new machine-readable digital document 200 from it. The text recognition module 107 is in Fig. 3 is shown schematically in detail, wherein it can be seen that it comprises an image optimization device 113, a separating and / or combining device 116, and a device 117 for optical character recognition. The image optimization device 113 is further subdivided into an alignment unit 114 and a contrast sharpness unit 115.

[0044] The digital document 200, which is classified as machine-unreadable, is first sent to the alignment unit 114 of the image optimization device 113, which aligns one or more pages of the digital document 200 by rotating them accordingly. Such rotation may be necessary, for example, if a scanned PDF document was fed into the scanning device and scanned at an angle or skew.

[0045] Once the alignment unit 114 has aligned all sides of the document accordingly, the thus-formed document is passed to the contrast sharpening unit 115 to sharpen or enhance the contrast accordingly. Sharpening or deepening the contrast may be necessary if, for example, the resolution during the scanning process is too low, the bit depth is too low (e.g., a 1-bit color depth instead of grayscale and color), a different type of poor contrast, or a noisy background.

[0046] Once the digital document 200 has been graphically optimized in the image optimization device 113, it is sent to the separating and / or combining device 116, which divides the digital document 200 into individual pages, so that each page is individually forwarded to the subsequent optical character recognition device 117. This "fragmentation" of the document and the subsequent recombining of the individual pages ensures that the page information of the digital document 200 is still present or retained after the optical character recognition.

[0047] The optical character recognition device 117 is configured to perform dictionary-based text recognition, resulting in improved text recognition. The language of the digital document 200 is also determined in advance, so that the dictionary corresponding to the language is also used for text recognition. Furthermore, the optical character recognition device (117) is assigned an artificial neural network that is configured and trained to recognize the correct word from the context despite individual unrecognizable letters. This artificial neural network has a recurrent circuit and an LTSM structure, which contributes to improved text / word recognition.

[0048] After the individual pages have been recombined by the separating and / or merging device 116, the digital, now machine-readable document is transferred to the search module 108. In the search module 108, the machine-readable digital document is searched for at least one keyword and / or at least one keyword modification generated based on a regular expression. The keywords or regular expressions to be searched are stored in the database 109 of the evaluation computer 104. The contents of the database 104 can be edited by a user, in particular an authorized user, of the provider computer 101, wherein, in particular, additional keywords and additional regular expressions can be stored in the database 109 in order to refine or improve future searches. Of course, it is also possible to modify existing entries in the database 109 or even delete them entirely.

[0049] An example of the contents of the database 109 is shown schematically in Fig. 4, which contains two keywords purely by way of example, in this case the keywords "garage" and "front door." Additionally, two different regular expressions are also included purely by way of example, which are "access\w*" and "(electronic / electrical)(-I)security(s)?" The search module 108 is designed to search the digital document 200 for modifications and variants of individual words or of related or compound terms using a predefined syntax. The syntax underlying the regular expressions is explained in more detail below using the exemplary contents of the database 109.

[0050] The individual keywords without the regular expressions only return hits in the digital document 200 that exactly match that keyword. The terms "front door" or "garage" would be found; however, "front door lock" or "garage door" would not be included in the search results. "Front door" would also not be included in the search results, so a deviation of just one letter would result in a word not being found. Therefore, there is the additional option of using "\w*" in the syntax, which returns hits for the keyword preceding this element with no additional letters, with a single additional letter, or with any number of additional letters appended to it. Thus, in the example shown, "access\w*" would return hits such as "access," "access door," "access control," etc.It is possible to use the syntax's "I", which represents a combination of terms in a logical "or". In the example, "(electronic / electrical)" would return hits containing either "electronic" or "electrical". In the example, the database entry is then followed by "(-I)", which means "-" (hyphen) or " " (space). It is possible to use the syntax's "()?" (contents of the brackets in parentheses with a question mark), which means "optional". In the example, "Fuse(s)?" would return hits containing both "Fuse" and "Fuse(s). The individual regular expressions can be combined in any way, so that, for example, "electrical fuse", "electronic-fuse(s)", and "electronic fuses" would be found among the hits. The syntax's "\d" stands for any digit between 0 and 9.The "\w" in the syntax stands for any single letter appended to the corresponding keyword that must appear in order to be included in the search results. The "\w+" in the syntax stands for one or more additional letters appended to the corresponding keyword that must appear in order to be included in the search results.

[0051] The search module 108 is configured to remember the locations where the keywords or keyword variations occur. The downstream output module 110 then generates a new digital output document 202 in which the retrieved keywords and keyword variations are highlighted or marked.

[0052] For example, the "comment function" in a PDF document can be used to indicate the position of a retrieved keyword or regular expression. The comment then contains the keyword or regular expression underlying the search. Additionally, the "highlight function" for PDF documents can be used to highlight or emphasize the retrieved hits. The digital output document 202 can, for example, correspond to the original format of the digital document 200, but supplemented with comments and / or highlights for the retrieved keywords and / or keyword variations. The user can then very quickly locate the relevant passages in the extensive document that are relevant for later creation and submission of a quote.

[0053] Alternatively or additionally, there is also the Fig. 2, the output module 110 generates an extraction document 201 that is smaller than the machine-readable digital document 200 and contains only the pages or areas relevant to the tender. Thus, less data transfer is required and the storage requirements are reduced.

[0054] Alternatively or additionally, the entire original output document 202, but with comments or highlights, can also be transferred to the provider computer 101. Sending can be done, for example, via email, although it is also possible to simply send an email notification (or a push notification via smartphone or smartwatch) about new output documents 202 available in the system portal. It is also possible for the output document 202 to be provided via another suitable interface, whereby provision in a document management system is also possible.

[0055] Additionally, the output module 110 can be configured to generate an output document 202 comprising a predefined structure, so that the formatting, preparation, and design of the output document 202 are always consistent, thus providing a simplified and familiar overview for a provider. In other words, the output document 202 is always provided in a uniform format, preferably as a PDF document.

[0056] Additionally, the output module 110 can be configured to return an XML document in machine-readable form, making it available for subsequent processing or for browsing by other providers of potentially different services. This enables subsequent machine processing (e.g., SAP import) of the extensive document processed in the evaluation computer 104.

[0057] Dashed is optional in Fig. 2 also shows a document generation module 118, which is designed to generate a predetermined document format from machine-readable typed documents before it is fed to the search module 108. The data structure for the search in the search module 108 is preset accordingly, even for machine-readable typed documents.

[0058] Based on Fig. 5 explains a computer program or the method according to the invention, which can also be executed natively on a computer.

[0059] First, a digital document 200 is provided to a computer or an evaluation computer 104 (step S1). Subsequently, the digital document 200 is typed by a typing module 106 to determine whether the digital document 200 is a machine-readable document or a machine-unreadable document (query S2). If the digital document 200 was typed as a machine-readable variant (+), the keyword and / or regular expression search is immediately performed in a search module 108 (step S8). If, however, a machine-unreadable digital document 200 is present, it is fed to an image optimization device 113 in a text recognition module 107 in order to graphically prepare the document in the image-based digital document format in advance. For this purpose, the image optimization device 113 comprises an alignment unit 114 which aligns one or more of the pages of the document 200 by means of a suitable rotation, if necessary.rotated by only a few degrees (step S3). The contrasts in the image are then enhanced and / or sharpened, for which purpose the image optimization device 113 has a corresponding contrast sharpening unit 115 (step S4). In order not to lose the page information, the document is divided into individual pages by means of a separating and / or combining device 116 (step S5), with each page being individually fed to an optical character recognition device 117 for text recognition (step S6). The individual pages recognized with text are combined again by the separating and / or combining device 116 into a common, now machine-readable digital document 200 (step S7).Subsequently, the machine-readable document is searched by the search module for at least one keyword relevant to the bid submission and / or for at least one keyword variation generated based on a regular expression (step S8). Subsequently, the computer or the evaluation computer 104 provides or outputs an output document 202 in which the corresponding keywords and / or keyword variations are highlighted and / or marked (step S9).

[0060] With the system, method, and computer program according to the invention, extensive text documents can be reviewed for their relevance to the tender submission, thus achieving efficiency and cost savings. The examined tenders or reduced-size extraction documents 201 can then also be stored in a memory of the evaluation computer 104 to allow future access. LIST OF REFERENCE SYMBOLS 100 systems 101 provider calculators 102 Transmitting and / or receiving module (provider computer) 103 Network 104 evaluation computers 105 Transmitting and / or receiving module (evaluation computer) 106 Typing module 107 Text recognition module 108 Search module 109 (Keyword / Regular Expression) Database 110 Output module 111 tender calculator 112 Sending and / or receiving module (tendering computer) 113 Image optimization facility 114 Alignment unit 115 Contrast sharpness unit 116 Separation and / or merging device 117 Optical character recognition device 118 Document generation module 200 digital documents 201 digital extraction document 202 digital output document

Claims

[1] System (100) for identifying and / or extracting information relevant for submitting a tender from a document relating to a tender or a request for proposals, with a network (103) which is designed to at least temporarily connect an evaluation computer (104) to at least one provider computer (101) for the exchange of digital documents (200, 201), with a transmitting and / or receiving module (102) assigned to the provider computer (101), which is designed to send digital documents (200) to the evaluation computer (104) via the network (103) or to retrieve such documents via the network (103), and with a transmitting and / or receiving module (105) assigned to the evaluation computer (104), which is designed to provide digital documents (200, 202) via the network (103) for retrieval by the provider computer (101) or to receive such documents via the network (103), wherein the evaluation computer (104) comprises: - a typing module (106) designed to type a digital document (200) passed to it as a text-based PDF document in the format of the extended markup language (XML), a text-based PDF document, a vector-based machine-readable document, or a raster-based machine-unreadable document, wherein the digital document (200) is first typed based on its file extension, and wherein, to avoid incorrect file extensions of the digital document (200), a preliminary analysis is carried out, in which, in the case of PDF documents, an attempt is made to extract text and only when a threshold value for the number of letters or characters is exceeded page by page or in the entire document is it concluded that it is a text-based machine-readable PDF document, wherein, if the threshold value is not reached, it is concluded that it is an image-based PDF document as a machine-unreadable document, - a text recognition module (107) which is designed to detect letters present in a digital document (200) previously typed as machine-unreadable and thus to generate a machine-readable digital document (200) from the machine-unreadable document, - a search module (108) designed to search the machine-readable digital document (200) for at least one keyword and / or for at least one keyword modification generated on the basis of a regular expression, wherein the at least one keyword and / or the at least one regular expression are stored in a database (109) of the evaluation computer (104), and - an output module (110) designed to generate a digital output document (202) in which the at least one found keyword and / or the at least one found keyword variation is highlighted and / or marked. [2] System according to claim 1,characterized by that the output document (202) is an extraction document (201) which is reduced in size compared to the machine-readable digital document (200), in which at least those pages are contained on which the at least one found keyword and / or the at least one found keyword variation is contained. [3] System (100) according to claim 1 or 2, characterized by that the machine-unreadable document (200) is present as an image-based portable document format, and that the text recognition module (107) has an image optimization device (113) for graphically pre-processing the document present in the image-based portable document format, which is connected upstream of a device (117) for optical character recognition. [4] System (100) according to claim 3, characterized bythat the image optimization device (113) has an alignment unit (114) which is designed to align one or more of the pages of the digital document (200) by means of rotation, and that the image optimization device (113) has a contrast sharpening unit (115) which is designed to strengthen and / or sharpen the existing contrasts. [5] System (100) according to claim 3 or 4, characterized by that a separating and / or combining device (116) is interposed between the image optimisation device (113) and the device (117) for optical character recognition, which is designed to provide the pages of the digital document (200) to the device (117) for optical character recognition in an individualised manner or to combine individual pages again to form the digital document (200). [6] System (100) according to one of claims 1 to 5, characterized byin that the search module (108) is designed to recognize text paragraphs or sections / chapters with identified keywords and / or identified keyword variations in the digital document (200), and in that the output module (110) is designed to output the pages of the digital document (200) with the at least one complete text paragraph or with the at least one complete section / chapter in the output document (202) in which the at least one found keyword and / or the at least one found keyword variation is contained. [7] A method for identifying and / or extracting information relevant for the submission of a tender from a document relating to a call for tenders or a request for proposals, comprising the steps of: - Providing a digital document (200) to a computer or an evaluation computer (104); - Typing the digital document (200) by means of a typing module (106) to determine whether the digital document (200) is in the format of the extended markup language (XML), a text-based PDF document, a vector-based machine-readable document, or a raster-based machine-unreadable document, wherein the digital document (200) is first typed based on its file extension, and wherein, to avoid incorrect file extensions of the digital document (200), a preliminary analysis is carried out, in which, in the case of PDF documents, an attempt is made to extract text and only when a threshold value for the number of letters or characters is exceeded page by page or in the entire document is it concluded that it is a text-based machine-readable PDF document, wherein, if the threshold value is not reached, it is concluded thatthat it is an image-based PDF document as a machine-unreadable document;, - creating a machine-readable digital document (200) by means of a text recognition module (107) if the document was previously typed as machine-unreadable; - searching the machine-readable document by means of a search module (108) for at least one keyword relevant to the offer and / or for at least one keyword modification generated on the basis of a regular expression, wherein the at least one keyword and / or the at least one regular expression is stored in a database (109) of the evaluation computer (104); - generating a digital output document (202) by means of an output module (110) in which the at least one found keyword and / or the at least one found keyword variation is highlighted and / or marked; and - Providing or outputting the output document (202) by the computer or the evaluation computer (104). [8] Method according to claim 7, characterized by that the machine-unreadable document (200) is in an image-based portable document format, that the text recognition module (107) comprises an image optimization device (113) in order to graphically prepare the document in the image-based portable document format in advance, wherein the image optimization device (113) comprises an alignment unit (114) which aligns one or more of the pages of the document (200) by means of rotation, wherein the image optimization device (113) has a contrast sharpening unit (115) which strengthens and / or sharpens the existing contrasts, and that the text recognition module (107) comprises a device (117) for optical character recognition which recognizes the letters of the previously graphically prepared document or a part thereof, wherein a machine-readable layer is added to the machine-unreadable digital document (200). [9] Method according to claim 8, characterized by that the pages of the digital document (200) are separated by means of a separating and / or merging device (116) before the optical character recognition. [10] A computer program loaded into a digital memory of a computer, which implements a method according to any one of claims 7 to 9.

Citation Information

Patent Citations

  • Color highlighting document image processing

    US20060062453A1

  • Automatic colorization of monochromatic printed documents

    US20080246998A1

  • Selective document retrieval method and system

    US5825943A