CN113282821A - Intelligent application prediction method, device and system based on high-dimensional session data fusion - Google Patents
Intelligent application prediction method, device and system based on high-dimensional session data fusion Download PDFInfo
- Publication number
- CN113282821A CN113282821A CN202110447927.XA CN202110447927A CN113282821A CN 113282821 A CN113282821 A CN 113282821A CN 202110447927 A CN202110447927 A CN 202110447927A CN 113282821 A CN113282821 A CN 113282821A
- Authority
- CN
- China
- Prior art keywords
- session
- model
- conversation
- neural network
- layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/216—Parsing using statistical methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an intelligent application prediction method, device and system based on high-dimensional session data fusion, wherein the method comprises the following steps: extracting features from the original data set to construct a session set; respectively inputting each conversation in the conversation set to a conversation-based embedded learning module, and outputting a group of conversation vectors; the session vector is input to a prediction model, from which the next application usage session is predicted. The method is based on the combination of the representation learning technology and the neural network model, and simultaneously introduces the global context information into the model, so that the available information of the model in the recommendation process and the model training process is increased, and the application prediction effect of the smart phone is improved.
Description
Technical Field
The invention particularly relates to an intelligent application prediction method, device and system based on high-dimensional session data fusion.
Background
At present, many smart phones have application program recommendation functions similar to application assistants, but the functions are often used for simply recommending a plurality of application programs recently used by a user according to the use history of the user, and the recommendation accuracy is poor. In addition, with the development of internet technology and the improvement of mobile phone storage space, the traditional application program representation mode may cause the problem of data sparsity, and the recommendation effect of smart phone application is seriously affected.
Disclosure of Invention
Aiming at the problems, the invention provides an intelligent application prediction method, device and system based on high-dimensional session data fusion, and the method, device and system are based on the combination of a session learning technology and a neural network model, so that the available information of the model in the recommendation process and the model training process is increased, and the intelligent application prediction effect is improved.
In order to achieve the technical purpose and achieve the technical effects, the invention is realized by the following technical scheme:
in a first aspect, the present invention provides an intelligent application prediction method based on high-dimensional session data fusion, including:
extracting features from the original data set to construct a session set;
respectively inputting each conversation in the conversation set to a conversation-based embedded learning module, and outputting a group of conversation vectors;
the session vector is input to a prediction model, from which the next application usage session is predicted.
Optionally, the method for constructing each session includes:
acquiring original data containing logs of different application programs accessed by a user;
extracting a user identifier U, an application identifier A and a time stamp T from the original data;
checking the timestamp T and deleting redundant data based on the checking result;
by using<U,A,T>Building a record, building a session using the record, the session s ═ a1,a2,…,ak,…,an) Application records defined as continuous use by a user over a period of time, akFor a particular application sequence of use, akBy<U,A,T>To construct.
Optionally, the embedded learning module is a neural network model for mapping heterogeneous sessions to fixed-length feature vectors based on a session-based embedding method, and an objective function of the embedded learning module is as follows:
wherein, N (a)i) Represents node aiP (c | f (a))i) Is observing a given node aiConditional probability of context neighborhood of (2):
wherein v isaAnd v'aRepresenting input and output vectors of the node a, wherein s is a session set, and sigma is a sigmoid function;
and a group of session vectors output by the neural network model are weight matrixes of the neural network model.
Optionally, the predictive model is a stacked recurrent neural network.
Optionally, the stacked recurrent neural network comprises a first layer of GRU models, a second layer of GRU models and a fully-connected output layer connected in sequence;
the first layer of GRU model and the second layer of GRU model are identical in structure and pass through the reset gate rtAnd an update gate ztTo convert the input, wherein:
rt=σ(Wrxt+Urht-1)
zt=σ(Wzxt+Uzht-1)
where σ is sigmoid function, xtAnd ht-1For the session vector and the preceding implicit state, W and U are learned weight matrices, htIs the current hidden state;
and the activation function of the fully-connected output layer is a sigmoid function.
In a second aspect, the present invention provides an intelligent application prediction apparatus based on high-dimensional session data fusion, including:
the construction unit is used for extracting features from the original data set and constructing a session set;
the embedded learning unit is used for respectively inputting each conversation in the conversation set to the embedded learning module based on the conversation and outputting a group of conversation vectors;
and the prediction unit is used for inputting the session vector into a prediction model, and predicting the next application program using session by the prediction model.
Optionally, the method for constructing each session includes:
acquiring original data containing logs of different application programs accessed by a user;
extracting a user identifier U, an application identifier A and a time stamp T from the original data;
checking the timestamp T and deleting redundant data based on the checking result;
by using<U,A,T>Building a record, building a session using the record, the session s ═ a1,a2,…,ak,…,an) Application records defined as continuous use by a user over a period of time, akFor a particular application sequence of use, akBy<U,A,T>To construct.
Optionally, the embedded learning module is a neural network model for mapping heterogeneous sessions to fixed-length feature vectors based on a session-based embedding method, and an objective function of the embedded learning module is as follows:
wherein, N (a)i) Represents node aiP (c | f (a))i) Is observing a given node aiConditional probability of context neighborhood of (2):
wherein v isaAnd v'aRepresenting input and output vectors of the node a as a session set, wherein sigma is a sigmoid function;
and a group of session vectors output by the neural network model are weight matrixes of the neural network model.
Optionally, the predictive model is a stacked recurrent neural network; the stacked recurrent neural network comprises a first layer of GRU model, a second layer of GRU model and a fully connected output layer which are connected in sequence;
the first layer of GRU model and the second layer of GRU model are identical in structure and pass through the reset gate rtAnd an update gate ztTo convert the input, wherein:
rt=σ(Wrxt+Urht-1)
zt=σ(Wzxt+Uzht-1)
where σ is sigmoid function, xtAnd ht-1For the session vector and the preceding implicit state, W and U are learned weight matrices, htIs the current hidden state;
and the activation function of the fully-connected output layer is a sigmoid function.
In a third aspect, the present invention provides an intelligent application prediction system based on high-dimensional session data fusion, including a storage medium and a processor;
the storage medium is used for storing instructions;
the processor is configured to operate in accordance with the instructions to perform the steps of the method according to any one of the first aspects.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides an intelligent application prediction method, device and system based on high-dimensional session data fusion, which are based on the combination of a session learning technology and a neural network model, and simultaneously introduce global context information into the model, thereby increasing the available information of the model in the recommendation process and the model training process, and further improving the intelligent application prediction effect.
Drawings
In order that the present disclosure may be more readily and clearly understood, reference is now made to the following detailed description of the present disclosure taken in conjunction with the accompanying drawings, in which:
FIG. 1 is a schematic flow chart of an intelligent application prediction method based on high-dimensional session data fusion according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a stacked recurrent neural network according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of the invention.
The following detailed description of the principles of the invention is provided in connection with the accompanying drawings.
Example 1
The embodiment of the invention provides an intelligent application prediction method based on high-dimensional session data fusion, which comprises the following steps as shown in figure 1:
(1) extracting features from the original data set to construct a session set;
(2) respectively inputting each conversation in the conversation set to a conversation-based embedded learning module, and outputting a group of conversation vectors;
(3) the session vector is input to a prediction model, from which the next application usage session is predicted.
In a specific implementation manner of the embodiment of the present invention, a method for constructing each session includes:
acquiring original data containing logs of different application programs accessed by a user;
extracting a user identifier U, an application identifier A and a time stamp T from the original data;
checking the timestamp T and deleting redundant data based on the checking result; in this step, some applications used by almost all users are deleted because they do not contain much useful information, like stop words in natural language, etc.;
by using<U,A,T>Building a record, building a session using the record, the session s ═ a1,a2,…,ak,…,an) Application records defined as continuous use by a user over a period of time, akFor a particular application sequence of use, akBy<U,A,T>To construct.
In a specific implementation of the embodiment of the present invention, since the session extracted from the dataset has a variable length and contains heterogeneous semantics and context information, in order to form a unified feature representation, a session-based embedding method is required to map the heterogeneous session to a fixed-length feature vector, and for this reason, in the embodiment of the present invention, the embedded learning module is a neural network model for mapping the heterogeneous session to a fixed-length feature vector based on the session-based embedding method.
In the session-based embedding problem, each application is treated as a node. If two nodes appear consecutively in the session, they are connected by a directed edge. In this way, the sequence required for the embedding phase is obtained.
In order to retain the structural information in the graph, the objective function of the neural network model in the embodiment of the present invention is to maximize the average logarithmic probability, and the expression is:
wherein, N (a)i) Represents node aiP (c | f (a))i) Is observing a given node aiConditional probability of context neighborhood of (2):
wherein v isaAnd va' input and output vector representations for node a; s is a session set, and sigma is a sigmoid function.
From equations (1) and (2) we can see that the computation of the gradient is very time consuming, since the gradient is proportional to the magnitude of | V |. This is impractical when faced with large-scale datasets. To reduce the cost of gradient calculations, we use the negative sample method. When training a node aiWhen we use their context nodes c e N (a)i) As positive samples and randomly selecting N nodes from the entire graph as negative samples, P (c | f (a))i) Can be expressed as:
wherein sigma is sigmoid function, and the optimization is carried out by adopting a random gradient descent method.
And a group of session vectors output by the neural network model are weight matrixes of the neural network model.
In a specific implementation manner of the embodiment of the present invention, the prediction model is a stacked recurrent neural network; the stacked recurrent neural network comprises a first layer of GRU model, a second layer of GRU model and a fully connected output layer which are connected in sequence;
the first layer of GRU model and the second layer of GRU model are identical in structure and pass through the reset gate rtAnd an update gate ztTo convert the input, wherein:
rt=σ(Wrxt+Urht-1)
zt=σ(Wzxt+Uzht-1)
where σ is sigmoid function, xtAnd ht-1For the session vector and the preceding implicit state, W and U are learned weight matrices, htIs the current hidden state;
the input to the GRU is the three-dimensional tensor of (samples, timepieces, input _ dim). In the prediction module, samples is the number of instances, timepieces is the number of required historical sessions, and input _ dim is the dimension of the session vector. To prevent overfitting, dropout layers need to be used after each layer. The hidden state of the last time step is used as the output of the GRU subnet. It will be further sent to a fully connected output layer which converts the hidden state into the final prediction session. The sigmoid function is chosen as the activation function, resulting in the final result, using binary _ cross as the loss function.
And the activation function of the fully-connected output layer is a sigmoid function.
Example 2
The embodiment of the invention provides an intelligent application prediction device based on high-dimensional session data fusion, which comprises:
the construction unit is used for extracting features from the original data set and constructing a session set;
the embedded learning unit is used for respectively inputting each conversation in the conversation set to the embedded learning module based on the conversation and outputting a group of conversation vectors;
and the prediction unit is used for inputting the session vector into a prediction model, and predicting the next application program using session by the prediction model.
In a specific implementation manner of the embodiment of the present invention, a method for constructing each session includes:
acquiring original data containing logs of different application programs accessed by a user;
extracting a user identifier U, an application identifier A and a time stamp T from the original data;
checking the timestamp T and deleting redundant data based on the checking result;
by using<U,A,T>Building a record, building a session using the record, the session s ═ a1,a2,…,ak,…,an) Application records defined as continuous use by a user over a period of time, akFor a particular application sequence of use, akBy<U,A,T>To construct.
In a specific implementation manner of the embodiment of the present invention, the embedded learning module is a neural network model, and is configured to map a heterogeneous session to a feature vector of a fixed length based on a session embedding method, where an objective function of the embedded learning module is:
wherein, N (a)i) Represents node aiP (c | f (a))i) Is observing a given node aiConditional probability of context neighborhood of (2):
wherein v isaAnd v'aInput and output vector representations for node a; s is the session set and σ is the sigmoid function.
And a group of session vectors output by the neural network model are weight matrixes of the neural network model.
In a specific implementation manner of the embodiment of the present invention, the prediction model is a stacked recurrent neural network; the stacked recurrent neural network comprises a first layer of GRU model, a second layer of GRU model and a fully connected output layer which are connected in sequence;
the first layer of GRU model and the second layer of GRU model are identical in structure and pass through the reset gate rtAnd an update gate ztTo convert the input, wherein:
rt=σ(Wrxt+Urht-1)
zt=σ(Wzxt+Uzht-1)
where σ is sigmoid function, xtAnd ht-1For the session vector and the preceding implicit state, W and U are learned weight matrices, htIs the current hidden state;
and the activation function of the fully-connected output layer is a sigmoid function.
The rest of the process was the same as in example 1.
Example 3
The embodiment of the invention provides an intelligent application prediction system based on high-dimensional session data fusion, which comprises a storage medium and a processor, wherein the storage medium is used for storing a plurality of session data;
the storage medium is used for storing instructions;
the processor is configured to operate in accordance with the instructions to perform the steps of the method according to any of embodiment 1.
The foregoing shows and describes the general principles and broad features of the present invention and advantages thereof. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (10)
1. An intelligent application prediction method based on high-dimensional session data fusion is characterized by comprising the following steps:
extracting features from the original data set to construct a session set;
respectively inputting each conversation in the conversation set to a conversation-based embedded learning module, and outputting a group of conversation vectors;
the session vector is input to a prediction model, from which the next application usage session is predicted.
2. The intelligent application prediction method based on high-dimensional session data fusion according to claim 1, wherein the construction method of each session comprises the following steps:
acquiring original data containing logs of different application programs accessed by a user;
extracting a user identifier U, an application identifier A and a time stamp T from the original data;
checking the timestamp T and deleting redundant data based on the checking result;
by using<U,A,T>Building a record, building a session using the record, the session s ═ a1,a2,…,ak,…,an) Application records defined as continuous use by a user over a period of time, akFor a particular application sequence of use, akBy<U,A,T>To construct.
3. The intelligent application prediction method based on high-dimensional session data fusion of claim 1, wherein the embedded learning module is a neural network model for mapping heterogeneous sessions to fixed-length feature vectors by a session-based embedding method, and the objective function is as follows:
wherein, N (a)i) Represents node aiP (c | f (a))i) Is observing a given node aiConditional probability of context neighborhood of (2):
wherein v isaAnd v'aRepresenting input and output vectors of a node a, wherein S is a session set, and sigma is a sigmoid function;
and a group of session vectors output by the neural network model are weight matrixes of the neural network model.
4. The intelligent application prediction method based on high-dimensional session data fusion according to claim 1, characterized in that: the predictive model is a stacked recurrent neural network.
5. The intelligent application prediction method based on high-dimensional session data fusion according to claim 4, characterized in that: the stacked recurrent neural network comprises a first layer of GRU model, a second layer of GRU model and a fully connected output layer which are connected in sequence;
the first layer of GRU model and the second layer of GRU model are identical in structure and pass through the reset gate rtAnd an update gate ztTo convert the input, wherein:
rt=σ(Wrxt+Urht-1)
zt=σ(Wzxt+Uzht-1)
where σ is sigmoid function, xtAnd ht-1For the session vector and the preceding implicit state, W and U are learned weight matrices, htIs the current hidden state;
and the activation function of the fully-connected output layer is a sigmoid function.
6. An intelligent application prediction device based on high-dimensional session data fusion is characterized by comprising:
the construction unit is used for extracting features from the original data set and constructing a session set;
the embedded learning unit is used for respectively inputting each conversation in the conversation set to the embedded learning module based on the conversation and outputting a group of conversation vectors;
and the prediction unit is used for inputting the session vector into a prediction model, and predicting the next application program using session by the prediction model.
7. The intelligent application prediction method based on high-dimensional session data fusion according to claim 6, characterized in that: the construction method of each conversation comprises the following steps:
acquiring original data containing logs of different application programs accessed by a user;
extracting a user identifier U, an application identifier A and a time stamp T from the original data;
checking the timestamp T and deleting redundant data based on the checking result;
by using<U,A,T>Building a record, building a session using the record, the session s ═ a1,a2,…,ak,…,an) Application records defined as continuous use by a user over a period of time, akFor a particular application sequence of use, akBy<U,A,T>To construct.
8. The intelligent application prediction method based on high-dimensional session data fusion according to claim 6, characterized in that: the embedded learning module is a neural network model and is used for mapping heterogeneous sessions to feature vectors with fixed lengths based on a session embedding method, and the target function of the embedded learning module is as follows:
wherein, N (a)i) Represents node aiP (c | f (a))i) Is observing a given node aiConditional probability of context neighborhood of (2):
wherein v isaAnd v'aRepresenting input and output vectors of a node a, wherein S is a session set, and sigma is a sigmoid function;
and a group of session vectors output by the neural network model are weight matrixes of the neural network model.
9. The intelligent application prediction method based on high-dimensional session data fusion according to claim 6, characterized in that: the prediction model is a stacked recurrent neural network; the stacked recurrent neural network comprises a first layer of GRU model, a second layer of GRU model and a fully connected output layer which are connected in sequence;
the first layer of GRU model and the second layer of GRU model are identical in structure and pass through the reset gate rtAnd an update gate ztTo convert the input, wherein:
rt=σ(Wrxt+Urht-1)
zt=σ(Wzxt+Uzht-1)
where σ is sigmoid function, xtAnd ht-1For the session vector and the preceding implicit state, W and U are learned weight matrices, htIs the current hidden state;
and the activation function of the fully-connected output layer is a sigmoid function.
10. An intelligent application prediction system based on high-dimensional session data fusion is characterized by comprising a storage medium and a processor;
the storage medium is used for storing instructions;
the processor is configured to operate in accordance with the instructions to perform the steps of the method according to any one of claims 1-5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110447927.XA CN113282821A (en) | 2021-04-25 | 2021-04-25 | Intelligent application prediction method, device and system based on high-dimensional session data fusion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110447927.XA CN113282821A (en) | 2021-04-25 | 2021-04-25 | Intelligent application prediction method, device and system based on high-dimensional session data fusion |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113282821A true CN113282821A (en) | 2021-08-20 |
Family
ID=77277320
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110447927.XA Withdrawn CN113282821A (en) | 2021-04-25 | 2021-04-25 | Intelligent application prediction method, device and system based on high-dimensional session data fusion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113282821A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117828280A (en) * | 2024-03-05 | 2024-04-05 | 山东新科建工消防工程有限公司 | Intelligent fire information acquisition and management method based on Internet of things |
-
2021
- 2021-04-25 CN CN202110447927.XA patent/CN113282821A/en not_active Withdrawn
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117828280A (en) * | 2024-03-05 | 2024-04-05 | 山东新科建工消防工程有限公司 | Intelligent fire information acquisition and management method based on Internet of things |
CN117828280B (en) * | 2024-03-05 | 2024-06-07 | 山东新科建工消防工程有限公司 | Intelligent fire information acquisition and management method based on Internet of things |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP3711000B1 (en) | Regularized neural network architecture search | |
CN109992773B (en) | Word vector training method, system, device and medium based on multi-task learning | |
CN111382868B (en) | Neural network structure searching method and device | |
CN116415654A (en) | Data processing method and related equipment | |
CN111709493B (en) | Object classification method, training device, object classification equipment and storage medium | |
CN113590900A (en) | Sequence recommendation method fusing dynamic knowledge maps | |
CN109961041B (en) | Video identification method and device and storage medium | |
CN109272332B (en) | Client loss prediction method based on recurrent neural network | |
CN113987147A (en) | Sample processing method and device | |
CN115688879A (en) | Intelligent customer service voice processing system and method based on knowledge graph | |
Wang et al. | NEWLSTM: An optimized long short-term memory language model for sequence prediction | |
CN111767697B (en) | Text processing method and device, computer equipment and storage medium | |
CN111357051A (en) | Speech emotion recognition method, intelligent device and computer readable storage medium | |
CN114692605A (en) | Keyword generation method and device fusing syntactic structure information | |
CN113362852A (en) | User attribute identification method and device | |
CN117350304B (en) | Multi-round dialogue context vector enhancement method and system | |
CN111310462A (en) | User attribute determination method, device, equipment and storage medium | |
CN114510576A (en) | Entity relationship extraction method based on BERT and BiGRU fusion attention mechanism | |
CN113282821A (en) | Intelligent application prediction method, device and system based on high-dimensional session data fusion | |
CN115525740A (en) | Method and device for generating dialogue response sentence, electronic equipment and storage medium | |
CN116738983A (en) | Word embedding method, device and equipment for performing financial field task processing by model | |
CN116957006A (en) | Training method, device, equipment, medium and program product of prediction model | |
CN116910190A (en) | Method, device and equipment for acquiring multi-task perception model and readable storage medium | |
US11941508B2 (en) | Dialog system with adaptive recurrent hopping and dual context encoding | |
CN111667028B (en) | Reliable negative sample determination method and related device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210820 |
|
WW01 | Invention patent application withdrawn after publication |