BE1025360B1 - MANAGING AND USING LEGAL FISCAL DOCUMENTS FROM LAW - Google Patents

Abstract

A computer-implemented method for sorting a collection of legal-tax documents from the case-law, whereby from this collection of legal-tax documents from the case-law a list of responsive documents is obtained, wherein the ordering of the list comprises a step in which each legal-tax document a weight parameter is attributed by making a summary that is tested against at least one criterion, these criteria being categorized.

Description

MANAGING AND USING JURI Dl SCH-FI SCALE DOCUMENTS FROM LAW

TECHNICAL DOMAIN N

The invention relates to a computer-implemented method for sorting a collection of legal-tax documents from case law, and a tangible non-transitory computer-readable information carrier comprising a computer program product for sorting a collection of legal-tax documents from the case law.

STAND OF THE TECHNIQUE

The Belgian legal system, as well as a number of other legal systems around the world, rely heavily on written legal advice and the written judgments of judges to explain or interpret laws on dispute resolution. As a result, judges and lawyers look for the most relevant past views or case law, thereby resolving or preventing disputes, while this amount of information continues to increase. Found cases are studied for relevance and are eventually quoted and discussed in documents, called litigation material, which for example advocate legal action, advise clients on probable legal actions or inform clients and lawyers about laws in certain jurisdictions.

In addition to the legal system, prior decisions or rulings are also relevant in Belgium. On the basis of these rulings, FPS Finance determines how tax laws will be applied to a specific situation or transaction that has not yet had an effect on tax matters.

EP2008 / 1 586 055 describes a method for managing and using information at law firms and potentially other companies. For example, one system provides a user interface for investigating case law for online legal investigation and for identifying and gaining access to law firm documents. Through the interface a user, such as a lawyer, can initiate or submit a legal investigation assignment and view the search results that not only identify relevant external documents of the online legal investigation service, but also relevant internal documents such as specifications, customer letters, and legal memoranda from the own collection of documents from the law firm.

The known methods, however, only focus on use by lawyers or other highly trained personnel, while access to legal-tax documents of relevant cases can also be of an informative and advisory nature for laymen.

BE2017 / 6017

Another problem with the way these legal-tax documents are organized is the lack of a user-friendly computer-implemented search terminal, where the results are ordered according to relevance to the search. The boolean search algorithm such as that according to Google is rather unsuitable for such an order, since the case law concerns a dynamic and discontinuous system of laws, judgments, judges and courts, whereby interpretations of laws may be linked to a spirit of time, social or social circumstances, different courts or judges, anticipations of future laws, and so on.

It is an object of the present invention to find a solution to at least some of the aforementioned problems.

SUMMARY OF THE INVENTION

In a first aspect, the present invention relates to a computer-implemented method for sorting a collection of legal-tax documents from the case law, according to claim 1.

In a second aspect, the present invention relates to a computer system for sorting a collection of legal-tax documents from the case-law, according to claim 12.

In a third aspect, the present invention relates to a tangible non-transitory computer-readable information carrier comprising a computer program product suitable for sorting a collection of legal-tax documents from the case law, according to claim 13.

The present invention is advantageous because a layman can, through the use of the computer system, without the intervention of a lawyer, inquire about legal-tax issues.

In addition, the invention is advantageous because the tax-legal documents are arranged according to current relevance.

DETAILED DESCRIPTION

The invention relates to a computer-implemented method and a tangible non-transitory computer-readable information carrier comprising a computer program product for sorting a collection of legal-tax documents from the case law. In the following, the invention is described in detail, preferred embodiments are explained, and the invention is illustrated on the basis of examples.

Unless defined otherwise, all terms used in the description of the invention, including technical and scientific terms, have the meaning as they are generally

BE2017 / 6017 are understood by those skilled in the art in the technical field of the invention. For a better assessment of the description of the invention, the following terms are explicitly explained.

"One", "de" and "it" refer to both the singular and the plural unless the context clearly presupposes otherwise. For example, "a segment" means one or more than one segment.

The terms "include," "include," "consist of," "include," "include," "include," "include," "include," and are inclusive or open terms indicating the presence of what follows, and which do not preclude or prevent the presence of other components, features, elements, members, steps, known from or described in the prior art.

The term "current relevance" or "current relevance" refers to the extent to which a particular document corresponds to the current dynamic and discontinuous system of laws, judgments, judges and courts, as interpretations of laws may be tied to a spirit of time, social or social circumstances, different courts or judges, anticipations of future laws, and so on. The current relevance of a document is determined in this invention by attributing its summary to a numerical value in the form of a weight parameter. This weight parameter is determined by a mathematical operation of several numerical weighting factors, with each weighting factor describing a certain criterion. The document whose summary has the largest weight parameter will therefore be the most relevant today. Related criteria are also preferably grouped into categories.

The term "classifier" refers to the abstract meta class of classification concepts that are used as a mechanism to collect elements with common properties and / or content in sets. In this document, this term refers to an algorithm designed to organize summaries and documents according to their characteristics.

The term "user" refers to a person who uses the content of data sets from an information system. In particular, the term in this document refers to the person who consults a content of the responsive legal-fiscal documents after passing instructions to the computer system.

The term "input device" refers to a device suitable for supplying input to a user device. This input is not limited to a certain modality and may include mechanical movement, sound, images, and the like. The input can be discreet and / or continuous. The input is also not limited by the number of degrees of freedom. The input can relate to both direct and indirect input. When providing input related to a position or its change, such as an indicator on a screen, the input can be absolute

BE2017 / 6017 as being relative. A non-exhaustive list of examples of input devices includes a keyboard, a computer mouse, a touch pad, a touchscreen, a camera, a scanner, a joystick, a microphone, a light pen, a trackball, a projected keyboard, and a game controller.

A "legal-tax document" refers to a written or electronic piece of evidence to make with legal aspects, or aspects relating to the law, of taxation.

The term "parsing" refers to parsing the grammatical structure of an entry according to a fixed grammar. A parser converts the entered text into a data structure similar to filling in a form with data at the specified places in a specified text format.

The term "server" refers to a computer system that provides services to users using the proposed computer-implemented method. The server preferably consists of at least one power supply, a memory, a processor, a hard disk, a motherboard, and a housing.

In a first aspect, the invention relates to a computer-implemented method for sorting a collection of legal-tax documents from the case-law, characterized in that a list of responsive documents is obtained from this collection of legal-tax documents from the case-law, arranging the list comprises steps whereby a weight parameter is attributed to each legal-fiscal document by making a summary which is tested against at least one criterion, these criteria being categorized.

To arrive at a summary of a document, each document goes through a number of steps. A summary is made for each document by extracting the text and metadata from the documents supplied, and then further processing them.

First of all, the plain text must be extracted from the documents. Plain text can be interpreted much better by a computer. Textual data exists in various forms such as HTML pages, PDF documents, Word documents, etc., each of which has its own form and format to upgrade the representation of flat text to something that can be used within their own environment. For example, with HTML different tags are added to enhance the layout or with PDF documents a fixed layout is made to make the document look the same under all circumstances. To be able to work with these different types of text within the proposed system, it is important to extract the plain text from these different document formats. In this way a common basis is created that will serve as input for further processing that will then no longer depend

BE2017 / 6017 is of the type of the document. In addition to extracting text, metadata is also collected. This data can be important when choosing the language-dependent mood algorithm to be used later in this step, or allows you to convert different types of document formats, such as pdf, doc, doex, csv, html, etc, into plain text without layout. The metadata can also include content that is important for classifying the documents, such as data, number of pages, and so on.

The document is then parsed, which comprises two parts. The first part is language detection and the second part is effective parsing of the text. The parser is chosen for each document that will produce the best result for that type. To obtain all the words used in a certain text, a tokenization process is applied. This process ensures that a text document is split into a stream of words by removing all punctuation marks and replacing all tabs and non-textual characters with spaces. The set of different words from all text documents is merged into the document collection dictionary. An NGram tokenizer is preferably used, since it gives the possibility to take two or three words together and to consider them as one word in the document collection. This can be advantageous with, among other things, person names, which often consist of two words that could be important for a text. Of course a proper name is also picked up if all words are considered separately. To reduce the size of the dictionary and therefore the dimensionality of the description of the document collection, the set of words is further reduced by applying filters or mood algorithms. Filters remove words from the dictionary and therefore from the summaries. For example, filtering can be applied to remove stop words that have little or no content. Examples are articles, connection words, prepositions, etc. Mood algorithms try to transform a word into the standard form of that word. They do this, for example, by converting plural nouns to the singular or by simplifying verbs to their root.

In order to better understand the whole of summaries, they must first be transformed into a characteristic vector that can then serve as input for a computer. The plain text in the document collection does not contain enough structure to be efficiently processed by a computer system. Despite the simple data structure, the vector space model ensures that large collections of documents can be analyzed efficiently. It represents documents as vectors in an m-dimensional space. Each document d from the document collection is described as a numerical characteristic vector w (d) = (x (d, ti), x (d, t _m )) where T = {ti, t _m } represents the dictionary. The main task of the vector space representation of documents is to find a suitable encoding of the characteristic vector. Each element of the vector usually represents a word (or group of words) from the document collection. The simplest way to encode a document is to use binary term vectors. If a word occurs in the document, the corresponding element is set to one,

BE2017 / 6017 does not occur, then it is zero. The encoding is thus reduced to a simple Boolean comparison. However, the importance of each word is considered to be equivalent. Weighting factors are used to improve performance. The weight assigned to a word is an indication of the importance or relevance of that word in a specific document or collection. A word with a high frequency in certain documents, but that does not or rarely occur in the entire document collection, is given a great weight. A weight w (d, t) for term t in document d is calculated as the term frequency tf (d, t) multiplied by the inverse document frequency idf (t) defined as idf (t) = log (N / nt). This describes the specificity of a specific term in a document collection. In addition to term frequency and inverse document frequency, a standardization is applied to ensure that all documents have the same chance of being found, regardless of their length. A more complex way to encode a document using binary term vectors is to use a more complex term vector that produces numeric values that can differ from 0 or 1. For example, an algorithm can be designed that the number of times a given word occurs in the document, but an even more complex algorithm can be used that uses the words to represent a certain interpretation of the summary in the form of a numerical value.

Attributes that are not relevant can also be removed by selecting on characteristics. In this way, a subset of the characteristics that are relevant to the target concept is sought. This ensures, among other things, that the data set becomes smaller, so that less computing power and search space is required for the effective processing of the vector model. It reduces the number of characteristics, removes irrelevant or redundant data and therefore ensures that the algorithms work faster. It also ensures, among other things, that the predicted accuracy and comprehensibility of the results is improved.

After preprocessing, the text is automatically placed in a structure of pre-designed categories, so that information can be obtained about texts with similar content. Texts from the same category will appeal to the user faster and may therefore be useful to offer together. Classification requires a previously known structure, which in this case is focused on legal-tax documents. A classifier can be trained to classify under certain categories based on a learning set. It is therefore the intention to choose these classifiers in such a way that a maximum number of summaries can be classified correctly for each category. To this end, an optimal classifier will be determined for each branch in the tree structure of a data set in order to arrive at a good and reliable result. The classification can also be done on the basis of the extracted metadata, or a combination of both methods.

This invention has the advantage that the method can be used to organize document lists of different types of searches. On the one hand, this method is possible

BE2017 / 6017 can be used for a categorical search, where the user gives instructions to search within a predetermined category by means of an input device. For example, the user can search by year, date or location by navigating to these categories based on the input device. On the other hand, this method can also be used for a search query in which the user submits as an instruction through an input device keywords to search the entire or partial collection of legally-tax documents for documents somewhat related to the entered keywords. It goes without saying that the present invention is not limited to the above types of searches. For example, both above-mentioned searches can be combined to start a second search based on a few keywords within a categorical search, or vice versa.

The present invention has the effect that through the use of the computer system by a layman, without the intervention of a lawyer or other highly trained personnel, he can better inform himself about legal-tax issues. The potential consequence of this is that the number of hours that can be invoiced by the lawyer can be reduced, which is reflected in a reduction in the cost price for the client and a potentially larger client portfolio for the lawyer.

In particular, the computer system can serve as an additional source of advice and information, all the more so because this method focuses specifically on a user-friendly and logical ranking of the legal-tax documents.

In a preferred embodiment, the aforementioned summary consists of at least one category, each category consisting of at least one criterion, with a numerical weighting factor attributed to each criterion. The value of the weighting factor indicates the current relevance of the corresponding criterion for the summary. The greater the numerical value of the weighting factor, the more relevant the relevant criterion is. The criterion itself is determined on the basis of a characteristic vector, which may or may not be a binary term vector.

In a further embodiment, the value of the numerical weight parameter of the summary is determined by performing a mathematical operation on the weighting factors. More preferably, this operation involves an addition.

In a further embodiment, the tax-legal documents are arranged according to decreasing value of the weight parameter of the summary in a list. This has the effect that the top document in the list is also the most current relevant document. All subsequent documents are therefore also ordered according to decreasing current relevance.

BE2017 / 6017

In a preferred form, this collection of legal-tax documents is logically grouped in a storage folder in a memory unit, thus facilitating the maintenance of the legal-tax documents.

In a further preferred form, this memory unit forms part of a computer system that is connected to the internet. This has the effect that the computer-implemented method can also be accessed remotely by users, for example by connecting to the computer system via a personal computer or smartphone.

Further preferably, this computer system is a server that has the advantage of being designed to manage, process, store and transmit data in a fully continuous system, whereby the computer-implemented method is potentially available 24 hours a day .

In a further preferred form, this collection of legal-tax documents is maintained by an authorized company. The effect of this is that the logical grouping of the legal-tax documents can be expanded, adjusted and / or maintained.

In one embodiment, the responsive documents contain information fields. This has the effect that the user can distinguish the individual documents from each other on the basis of this information, thus making the search more user-friendly. These information fields preferably comprise titles, the article number, the court and the date concerning the document.

In one embodiment, the responsive documents include action fields that give the user access to the content of the document. This is preferably an action field that calls up a short summary, and / or an action field that calls up the entire content of the document.

In a further embodiment, the short summary of a responsive document is preferably characterized by a title, an article, a court, a date, a book, and one or more tags.

In a further embodiment, the aforementioned tag refers to a keyword common in the collection of legal-tax documents. In a preferred form this tag forms an action field which initiates a search according to the relevant keyword.

In a second aspect, the invention relates to a computer system for sorting a collection of legal-tax documents from the case-law according to claims 1-11, characterized in that the computer system is configured for:

BE2017 / 6017 (a) receiving legal-tax documents via at least one unom unification network;

(b) storing these documents in a local non-transitory computer-readable information carrier;

(c) obtaining one or more keywords from the user via at least one communication network in the form of a search instruction;

(d) selecting a source group with tax-legal documents;

(e) identifying search information contained in the source group (f) creating an ordered results list based on the user's instructions from step (c) according to step (e);

(g) storing the ordered results list in a local memory;

(h) retrieving the ordered results list stored in the local memory.

(i) sending via at least one communication network the ordered results list to the user's computer system from step (c).

In a third aspect the invention relates to a tangible non-transitory computer-readable information carrier comprising an computer program product suitable for sorting a collection of legal-tax documents from the case law according to claims 1-11, the computer program product comprising instructions for:

(a) receiving legal-tax documents via at least one communication network;

(b) storing these documents in a local non-transistory computer-readable information carrier;

(c) obtaining instructions from the user via at least one communication network;

(d) selecting a source group with tax-legal documents;

(e) identifying information contained in the source group;

(f) creating an ordered list based on the user's instructions from step (c) according to step (e);

(g) storing the ordered list in a local memory;

(h) retrieving the ordered list stored in the local memory.

(i) transmitting, via at least one communication network, the ordered list to the user's computer system from step (c).

In the following, the invention is described a.d.h.v. non-limiting examples illustrating the invention, and which are not intended or may be interpreted to limit the scope of the invention.

EXAMPLE

The invention will now be further elucidated with reference to the following example, without however being limited thereto.

BE2017 / 6017

The following example concerns an embodiment in which the summary is drawn up on the basis of the categories, criteria and weighting factors as shown in Table

1. The criteria in this case correspond to the aforementioned characteristic vectors, binary or otherwise.

The summary according to table 1 contains the five categories "judge", "law", "time", "content" and "specificity". Each category consists of different subcategories, called criteria here.

The "judge" category includes various criteria that relate to the judge who produced the court ruling in that document. Hereby the content of the document is checked on the name of the judge, his reputation, whether the ruling brought about changes in the case law, whether the ruling included a precedent, and whether the ruling took place in a higher court such as the Court of Appeal or Court of Cassation.

The "law" category includes various criteria that relate to the legislation mentioned in the document. The content of the document is checked to see if there are changes in legislation, whether a certain legislation was applicable at the time of the judgment, and whether a certain legislation was applicable at the time of the facts.

The "time" category includes various criteria that relate to data and duration. Hereby the content of the document is checked on the date of the plea, the date of the decision, the date of the facts, the date of dispute, and the time between the facts and decision.

The "content" category includes various criteria that relate to the general content. Here the content of the document is checked for the presence of references to other legal literature, references to other case law, the content of the judgment and its novelty, as well as the number of pages, the number of conflicting judgments, the number of judgments cited from the past, and the number of statements in the same matter.

The "specificity" category includes various criteria that relate to the novelty of the judgments. The content of the document is checked for possible innovative passages and precedents.

TABLE 1

CATEGORI E WEGI NGS FACTOR CRITERI UM Judge 1 Does the document contain a change in case law? 1 Does the document include the name of the judge?

BE2017 / 6017

2 Does the document include judgments from higher courts? 2 Is the ruling innovative? Does the judgment have a certain topical value? 2 Does the court have a known reputation? 3 Are precedents known in case law? Law 1 Does the document contain a change in legislation? 2 Does the document include the legislation applicable at the time of the judgment? 3 Does the document include the legislation applicable at the time of the facts? Time 1 Does the document include the date of the plea? 1 Does the document cover the time between the facts and the ruling? 2 Does the document include the date of the ruling? 2 Does the document include the date of the facts? 2 Does the document include the date of the dispute? Content 1 Includes references to other legal literature 1 Includes references to other case law 1 How many pages does the pronunciation include? After all, this is a criterion that points to the thoroughness of the analysis. 1 Have there been multiple conflicting statements about the same subject? After all, it can be an added value that this point is clarified in order to achieve certain legal certainty. 2 How many judgments by the same judge have been delivered in the past? After all, this is less relevant the more pronounced judgments there are. 2 How many statements in the same matter have been made in the past? After all, this is less relevant the more pronounced judgments there are. 3 To what extent can the content of the statement and the innovative passage be used in other cases? Specificity 1 Is there an innovative passage in the statement? 1 Does the document refer to a precedent?

BE2017 / 6017

Claims (13)

CONCLUSIONS A computer-implemented method for sorting a collection of legal-tax documents from the case-law, characterized in that from this collection of legal-tax documents from the case-law a list of responsive documents is obtained, wherein the ordering of the list includes a step in which a weight parameter is attributed to each legal-fiscal document by making a summary that is tested against at least one criterion, these criteria being categorized and summaries being made, first being transformed into a characteristic vector that can then serve as input for a computer, which summaries are represented as vectors in an m-dimensional space, each document d from the document collection being described as a numerical characteristic vector w (d) = (x (d, ti), x (d, tm)) where T = {ti, t _m } represents a dictionary. A computer-implemented method according to claim 1, characterized in that each category consists of at least one criterion, with a weighting factor attributed to each criterion. A computer-implemented method according to claim 2, characterized in that the weight parameter is determined by performing a mathematical operation on the weighting factors. A computer-implemented method according to claim 3, characterized in that tax-legal documents are ordered in the results list according to decreasing value of the weight parameter. A computer-implemented method according to claim 1, characterized in that this collection of legal-tax documents are logically grouped in a storage folder in a memory unit. A computer-implemented method according to claim 5, characterized in that this memory unit forms part of a computer system connected to the internet. A computer-implemented method according to claim 1, characterized in that this collection of legal-tax documents is maintained by an authorized company. A computer-implemented method according to claim 1, characterized in that the responsive documents contain information fields such as preferably title, article, court and date. A computer-implemented method according to claim 8, characterized in that the responsive documents contain action fields such as preferably an action field that calls up a short summary, or an action field that calls up the entire contents of the document. A computer-implemented method according to claim 9, characterized in that the short summary of a responsive document is preferably characterized by a title, an article, a court, a date, a book, and one or more tags. BE2017 / 6017 A computer-implemented method according to claim 10, characterized in that the aforementioned tag refers to a keyword of the document. A computer system for sorting a collection of legal-tax documents from the case law according to claims 1-11, characterized in that the computer system is configured for: receiving legal-tax documents via at least one communication network; b. storing these documents in a local non-transitory computer-readable information carrier; c. obtaining instructions from the user via at least one communication network; d. selecting a source group with tax-legal documents; e. identifying information contained in the source group; f. creating an ordered list based on the user's instructions from step (c) according to step (e); g. storing the ordered list in a local memory; h. retrieving the ordered list stored in the local memory. i. transmitting via the at least one communication network the ordered list to the user's computer system from step (c). A tangible non-transitory computer-readable information carrier comprising a computer program product suitable for sorting a collection of legal-tax documents from the case law according to claims 1-11, the computer program product comprising instructions for: receiving legal-tax documents via at least one communication network; b. storing these documents in a local non-transitory computer-readable information carrier; c. obtaining instructions from the user via at least one communication network; d. selecting a source group with tax-legal documents; e. identifying information contained in the source group; f. creating an ordered list based on the user's instructions from step (c) according to step (e); g. storing the ordered list in a local memory; h. retrieving the ordered list stored in the local memory. i. transmitting via the at least one communication network the ordered list to the user's computer system from step (c).